← Back to leaderboard
datachaz

Charly Wargnier ♨️

Verifiedactive
@datachaz

Deep in Agents, LLMs and automation · Open source at ♥ · Sharing AI gems daily to help you work smarter · Ex-data scientist @Snowflake | @Streamlit ·

RosterAI Model News & ReleasesAI Agents & Productivity Tools
182,455 followers · 49,391 following · Last polled 3h ago
Perf. Score
60.2
Rank #144
Value Score
38.8
#169 of 273 priced
Median views
9.9K
p25 4.4K · trailing 7d
Followers
182K
+0.1% 7d
ER (impr.)
1.21%
eng ÷ impressions
ER (foll.)
0.18%
eng ÷ followers
Est. CPM
$35.32
QT-rate basis
WoW views
+24.9%
▲ rising

Rates & value · est. from 9.9K median organic views

fairly priced
Quote tweetbasis
$350
$35.32 / 1K views
Post
—
no CPM
Thread
—
no CPM
Retweet
—
no CPM
Value Score 38.8 · rank #169 of 273 priced$4729.73 / 1K engagementsrate age unknown (pre-audit import)

Trends

Follower growth
Median views / day
Engagement rate
View distribution · trailing-7d posts

Each blue dot is one organic post. The Performance Score ranks by the median (9.9K), not the mean (27K) — so a single viral post doesn't inflate a creator's reach. The p25 floor (4.4K) is the level 75% of posts clear.

Recent posts · ranked by views

PostPostedViewsLikesRepostsRepliesQuotesBookmarksER
.@grok is this real? https://t.co/WDzWpceRyafrozen19d ago1.8M2.8K94733346600.24%
probably the best way to visualize the continuous progress in humanity https://t.co/NIrUbFREehfrozen6d ago831K6.8K1.1K1671151.3K1.15%
OpenAI employees finding out who owns https://t.co/WqgHr1bXny https://t.co/cpmkYlS5ZT
frozen
23d ago
599K
2.8K
100
80
20
432
0.57%
Training your own Jev in minutes. For $17. https://t.co/40SfsSb1Th29d ago510K2.4K15737103.9K1.28%
..and they say romance is dead https://t.co/if8lYznTorfrozen28d ago476K16K1.3K1041031.4K3.96%
HOLY F*CK, THIS GUY LITERALLY REIMPLEMENTED ALL ADOBE PRODUCTS FROM SCRATCH, FREE AND OPEN SOURCE 🤯 > Photoshop → PhotoCraft > Illustrator → VectorCraft > Premiere Pro → FilmCraft > Lightroom → LightCraft > Acrobat → PdfCraft > After Effects → EffectCraft > InDesign → DesignCraft Free. Open source. Built in Rust. link in 🧵 ↓frozen15d ago335K8.9K1.3K150908.9K5.79%
FINALLY an AI that outsources to humans. We're so f*cking back. https://t.co/Dgu4QlcxkZfrozen24d ago299K962541423430.46%
silly question: if Opus 5.5 is better than Fable 5.1, why would we ever use Fable 5.1 again?frozen25d ago290K1.4K32319800.58%
US AI labs: "China isn't even close." US AI labs: https://t.co/4ztqlbGLmAfrozen13d ago251K3.4K144102242241.54%
vibe coders these days: https://t.co/Z2ZqnKNhZlfrozen16d ago221K2.7K24149347511.71%
this is exactly how I imagined a web crawler would look https://t.co/ORAd2VsPJNfrozen21d ago199K4.9K573148681.2K3.43%
I CAN'T STOP THINKING ABOUT THIS ONE An AI that watches your face, picks up on hesitation, and jumps in at the right moment. All while it’s talking. @heytavus's new Griffin fooled 48% of people on a live call. Honestly, I'm not surprised. https://t.co/Lw1Nv5QRxU21d ago160K3551222532850.49%
a bride just gifted her husband a $12,000 GPU. and they say romance is dead. https://t.co/kF38kOgVA9frozen6d ago157K1.1K10562472070.95%
My bro Dario has Mr Bean vibes 🤣 https://t.co/NvzLSoHFl8frozen22d ago140K1.6K8074292501.47%
THIS CAME OUT WAY TOO GOOD NOT TO TURN INTO A TUTORIAL The whole video was made with @Pexoai_offical, and the workflow is honestly pretty slick. Instead of endlessly tweaking prompts and hoping for a better result, you can just work through the idea with it conversationally. Then it starts putting the whole thing together: > builds the storyboard first > lets you tweak individual scenes > fixes specific parts without rerendering everything That makes the process feel way less random and much more like actually directing the video. If you’ve been curious about @Pexoai_offical, the tutorial below walks through the full thing. Watch to the end, then check the replies for the finished video 👇frozen27d ago123K9159330430.18%
COMPANIES WILL SPEND $2.5 TRILLION ON AI THIS YEAR (source: Gartner) ...so naturally, everyone built a gateway. Cloudflare. Vercel. Portkey. Kong. LiteLLM. OpenRouter. They route models. CData is trying to solve what happens before the model even gets the prompt: → 350+ live systems → real company definitions → persistent business memory Its Connect AI layer hit 98.5% accuracy across 378 prompts. That's 25+ points ahead of other MCP providers. Definitely a company worth keeping an eye on. 👀frozen23d ago114K76215040.09%
Opus 5.5 users switching back to Opus 5.5 after trying another model https://t.co/5ftFbFt7TPfrozen25d ago112K4.3K817013934.09%
🚨 BREAKING: vibe-coders have announced that databases are no longer needed: https://t.co/bH5V7RmiLDfrozen26d ago107K99867237421101.36%
OPUS 5.5 IS REALLY A DIFFERENT BEAST 🤯 In @aimlapi’s latest test, 4 of the hottest models get locked in a tiny box with a Rubik’s Cube. > Jev and Laya fire off 20 random moves in seconds > GPT-6 Sol makes one bad choice and gives up > Opus 5.5 waits a minute… then solves it in 👏 three 👏 moves ↓frozen27d ago104K24250292910.40%
that plaque definitely needed an upgrade 😁 https://t.co/6lIT4CI0kHfrozen12d ago100K3.6K26727193334.23%
Sorry, son. Dad is busy burning tokens. https://t.co/DERQ0cP9Yifrozen28d ago93K1.9K14363291352.40%
most companies adding AI to their products: https://t.co/cLpyoZusa6frozen27d ago89K3.6K38770422334.89%
i don't think people realize how big this is… A US lab just released a 40B open-weight coding model leading its class on three major coding benchmarks. 125+ tok/s on a MacBook. 150+ on RTX Spark. no per-token fees. no usage limits. your laptop is now the endpoint. https://t.co/xQJXBhUJRi1d ago89K724824837091.77%
AI is getting out of hand https://t.co/kbvrzHvT7Hfrozen10d ago88K84412518213621.56%
THIS IS GETTING SERIOUSLY INTERESTING @DynaRobotics' Dyna-2.1 is the first Physical Agent capable of completing an entire job end-to-end. It reasons across long-horizon workflows, dynamically chaining multiple physical tasks into a single autonomous process 🤯 all details below ↓frozen23d ago85K8421110140.15%
me when people ask about the current state of AI: https://t.co/lMFLgZjTvF12h ago79K91213736133751.87%
Claude's just had enough https://t.co/IJNHGouENSfrozen20d ago74K2.9K11765151314.29%
We need more data centers https://t.co/jnCdl04Cwlfrozen17d ago66K1.3K16242273582.86%
Gen Z kids through the years https://t.co/Au87wUhC25frozen4d ago65K3.6K20749142996.34%
YOUR AI AGENTS SHIP FAST. BUGS DO TOO 🐛 @sazabi just shipped and it can watch your logs, detect anomalies and send agents to investigate. The goal → catch real issues before users notice. Love this angle. Faster shipping needs faster answers ↓frozen9d ago61K39155040.10%
✅ Claude Code ✅ Claude Cowork ✅ Claude Design ✅ Claude Science ✅ Claude Finance ✅ Claude Legal ✅ Claude Teacher ✅ Claude Pharma ⬜ Claude HR ⬜ Claude Analytics ⬜ Claude Marketing ⬜ Claude Sales ⬜ Claude R&D ⬜ Claude Accounting ⬜ Claude Engineering https://t.co/gggIwUiVuGfrozen29d ago57K42342811091.02%
man I'm never ever leaving this app 😁 https://t.co/rM80C80p3gfrozen26d ago53K87023141431.78%
MEET YOUR AI MARKETING DEPARTMENT. → Strategist researches your niche → Creator makes videos → Analyst tracks conversations → Manager plans and publishes Freaking. cool. That's Muse from @Adant_ai. Four roles, one team, working 24/7 👇frozen9d ago49K391770170.16%
nailed it https://t.co/NJBQUZTVMQfrozen25d ago48K508302710601.33%
WHAT?! Unreal Labs published an awesome playbook on building cloud agents 1M+ agents launched. One year of battle-testing. @unreallabsai breaks down the stack, security choices and integration headaches. super practical lessons for anyone taking agents into production 👀frozen9d ago47K462540100.18%
Tableau and PowerBI analysts watching this launch: https://t.co/IswrJTZcLofrozen12d ago45K223181021010.78%
i feel seen 😅 https://t.co/TwvfbzvgSmfrozen21d ago45K41835198861.26%
WOW Base Code can turn any GitHub repo into a ready-to-use cloud environment 🤯 Infra, database, services, previews → already set up before you start. And it all runs in the browser. Zero local setup ↓frozen24d ago43K441150270.20%
"How browsers work" in 40 seconds. Opus 5.5 drew every single frame of this animation 🤯 https://t.co/tXTDoQSAtXfrozen29d ago42K472252043232.02%
WE'RE ENTERING A NEW ERA OF INTERFACES a smart ring that doesn’t make me stare at another screen but still gives me access to Muse and all my agents waitlist is open. link in 🧵↓frozen14d ago41K611860160.25%
this guy created Linux without Opus 5.5 https://t.co/ekMcjVPerjfrozen8d ago41K2.5K1147721786.95%
Last week, I tuned into the @Atlassian State of AI SDLC Digital Summit and something stopped me: 94% of engineering teams have adopted AI. Only 6% have the systems in place to scale a truly AI-Native SDLC. Mike Cannon-Brookes (CEO + Co-Founder at Atlassian) sat down with Guillermo Rauch (CEO of @Vercel) to unpack exactly how to close this gap, and their conversation was one of those sessions where you pause and rewind. The core thesis? Competitive advantage won't come from just adopting AI, but from how well you connect context, including projects, owners, dependencies, and decisions, to the agents doing the work. That framing ran through the entire summit, and it's what made the lineup genuinely worth the time: 🎯 Tamar Yehoshua (Atlassian) and Elena Verna (Lovable) on the role of judgement: "When you can develop so fast, you have to pick what you're developing and make sure you're doing the right thing." 📊 Justin Reock (DX) and Uma Namasivayam (@Dropbox) on why PR volume and token spend aren’t the only metrics you should be paying attention to, and what to measure instead. ⚙️ Matt Canham and Ming Wu on scaling to 1M agents across 6K engineers at Atlassian, with a line I keep coming back to: "The agent is interchangeable. The system of record and the context underneath it are not." If you're serious about moving from AI adoption to actual delivery, there’s a lot worth digging into here. Missed a session, or want to revisit one? They’re all available on demand now → https://t.co/rKYJSKH7wEfrozen16d ago41K51243160.21%
HALF MY REPLIES LAND WHILE I'M ASLEEP I post from Europe. A big part of you reads me from the US. So the questions come in at 1 AM: → does it run locally? → is it actually free? → what's the catch? By the time I wake up they're 7 hours old. I'd answer a few, and the rest just sat there. I tried batching them every morning. It ate the first hour of my day and I still missed half. So I started using an AI assistant in Slack to help review questions overnight. Last night, while I slept, it: → reviewed 163 replies across my last 9 posts → identified 29 genuine questions → checked suggested answers against the repo's docs → prepared 29 draft responses for my review → flagged 4 where the replier was right and I was wrong Nothing is posted automatically. At 8 AM, I review the drafts, make any necessary edits and publish the replies myself. 29 of 29 answered before 9 AM. Try free at @viktor_com. $100 in credits, no card. Full link in my first reply.frozen7d ago40K28113010.11%
OPUS 5.5 JUST COOKED GPT-6 SOL 🤯 Same Blender brief Same 3D pelican VERY different vibes. try both on https://t.co/DfWaNXYTayfrozen1mo ago39K761791110.29%
V4 TURBO FINALLY MADE OUR AGENTS SOUND RIGHT 🤘 Stoked for ElevenAgents Architect to keep building ↓ https://t.co/m6zVrWBFxwfrozen16d ago39K781831150.30%
Cursor made it easier to go from an idea to working code. 'Syren Video' is taking that approach to video: prompt to create → chat to refine. now THIS is the "Cursor for video" vision I’m here for 👀frozen13d ago38K371920110.18%
YOUR AGENT’S PRODUCTION TRACES ARE NOW TRAINING MATERIAL So cool! @OvermindLab now connects your code and traces, builds a context graph, and turns real interactions into training and eval datasets. Use them to compare prompts and models, investigate failures, and fine-tune a smaller model on your product’s own data. Best part? It’s open source, you can self-host it, and you own the weights 🤘 repo link in 🧵↓frozen16d ago37K5623180210.32%
finally, something that isn’t AI https://t.co/D6pRdj6tyufrozen26d ago34K22812245430.91%
MUSE AND INSTINCT GET THE SPOTLIGHT. PHOTON POWERS THE MESSAGING BEHIND THEM. @PhotonHQ just raised a $4.5M seed round. Its open-source framework, Photon , connects AI agents to iMessage, WhatsApp, Telegram and more. Some crazy traction since April! → 450K monthly npm downloads → 50K+ developers → 10x revenue growth Building your own personal agent? Photon gives you a messaging layer to build on, with rich media, group conversations, cross-channel routing and fallbacks. The infra behind those consumer experiences could power yours too! Free, MIT-licensed and open-source. Repo link in 🧵↓frozen18d ago34K45188180.24%
An open-source alternative to Palantir Foundry is here. And it looks seriously exciting! Meet Helix Foundry! It connects your databases, APIs, files, and services like Stripe into a company ontology: a graph linking customers, transactions, users, and other business entities. You can query it, explore those relationships, and ask questions using AI, with answers backed by executed SQL. Under the hood: → HelixDB stores the ontology → S3-compatible storage holds data snapshots → DuckDB transforms and analyzes the data Very excited to see this kind of tooling go open source. Repo link in 🧵↓frozen13d ago33K396382318493.93%
You know what's annoying when you're deep into an AI workflow? Remembering something you need to follow up on, then having to stop everything just to write it down. @SpeakON_Global gives you an actual physical button to capture that thought by voice, with an AI agent to help with the next steps. Pretty neat2d ago32K1937030.10%
accurate af https://t.co/m5WQgQVzAWfrozen13d ago32K52546319822.17%
Just got my new MacBook 👀 What should I install first? https://t.co/dbx6nfZ80l17d ago32K276331811141.97%
🚨 @OpenBMB just dropped MiniCPM-V 4.7, and the engineering behind it is seriously impressive. 35B total parameters, ~3B active at inference, and ~3GB memory in its target on-device configuration. Here's what makes it work: → LES predicts which experts the next Transformer layer needs and preloads their weights to reduce memory latency. Misses get rescheduled to maintain parallelism. → DSpark Speculative Decoding drafts tokens ahead of the main model, which verifies them. Same output, faster execution. → Visual Token Compression cuts image and video token overhead by 4x and 16x. That last one makes a real difference. So I built something to test it 👇 🎬 Clip Narrator: drop in ANY video clip and get back a fully narrated version. My test footage? Literally 10 seconds of milk heating in a pot. The 35B-A3B sampled 8 frames, compressed the clip into ~2,500 visual tokens, and generated a timestamped movie-trailer script: "A pot of milk heats up." "Steam rises. The stakes are high." "It boils over! Disaster strikes!" A TTS model brought it to life, generating 8 seconds of audio in just 2.6 seconds. Subtitles synced automatically. Export and done. Here's what impressed me: it wasn't just captioning 8 still frames. It tracked the milk rising, steam building, and boil-over, timing each line to the action. That's temporal reasoning, not just image description! A 1.3B variant also handles OCR, charts, math, images and video. And both support 256K context, with fully open-source weights, code and fine-tuning pipelines. Beyond scaling parameters, it's about smarter execution on consumer hardware. Try it locally and tell me the most boring clip you'd feed it first 👇 FREE and open-source. Repo link in 🧵↓frozen6d ago31K33136140.18%
MY PREDICTION FOR 2028: Most creatives will be managing teams of AI agents before they ever hire another human. Crazy? Maybe. But I can see it happening. What's your prediction? Share it in @contraben's thread🧵 and you could win a MacBook Pro 👇 https://t.co/c9chqN3CKFfrozen14d ago31K43204270.25%
ANTHROPIC JUST REVEALED HOW CLOSE THEY ARE TO RECURSIVE SELF IMPROVEMENT The transparency around their internal R&D acceleration is staggering. AI agents now handle 26% of all research at Anthropic. That number was under 1% just six months ago 🤯 The oversight metrics are equally massive → 30,000 autonomous agents running concurrently → Over a billion decisions reviewed last month → Only 50 weekly transcripts escalated to human review They explicitly stated this data shows how close the world is to RSIfrozen28d ago30K15831143810.95%
JEV FOUNDER DIOGO ALMEIDA (@CompleteSkeptic) JUST DROPPED A PDF ON BUILDING A JEV HARNESS FOR CODING AGENTS It's a great blueprint for making your coding agents 200× faster and 400× cheaper. Link to the 12-page PDF in the 🧵 ↓ Also bookmark @zodchiii’s ace article on Jev. Well worth the read. now drop this PDF + article into Codex or Claude and put it to work ;)frozen29d ago30K187231513071.77%
🚨 ORACLE AND @DEEPLEARNINGAI JUST DROPPED A MASTERCLASS ON ADAPTIVE AGENTS. SO MUCH value packed into one hour, and it's completely FREE 🤯 The harsh reality right now is that AI agents forget everything when a session ends, forcing you to pay for the exact same mistakes every. single. day. Nacho Martínez and Casius Lee teamed up with @AndrewYNg and @DeepLearningAI and to completely dismantle this problem. The pacing of this curriculum is what makes it great. In just over an hour, it teaches you how to architect 3 heavy adaptation layers from scratch: ▪ Part 1 - Behavior: → you will convert agent traces into reusable skills, setting up a human approval gate to ensure only verified logic moves forward. ▪ Part 2 - Knowledge: → you will map your codebase by building a knowledge graph from imports and historical Git co-edits, proving exactly why graph retrieval beats basic keyword search. ▪ Part 3 - Model: → you will learn to identify the exact thresholds where fine-tuning and weight-space methods become necessary for your stack. You will leave with a complete understanding of how to make agents measurably better with every run. link to free course in 🧵 ↓frozen29d ago30K532382240.37%
THIS IS WILD. HEADLESS SAAS, OPERATED BY PROTO. > Describe what you need > https://t.co/qmTJYLAo6k builds the system > Proto runs it No per-user fee for manual work. Agent reads, writes and workflows use credits. gotta love @protosphinx's take on Headless SaaS 💥 https://t.co/REUQwI5QxUfrozen7d ago30K43201130.23%
Building open-source AI in Europe? This one's for you! @nebiusai really cooked with its European AI Builder Program expansion 🔥 $10M across compute grants, capital and community. Compute grants support the tools builders need, from open models and agent runtimes to evals and inference tooling. Europe has the talent. Now let's give promising projects room to grow 👊frozen4d ago29K36148040.21%
THIS IS FREAKING DANGEROUS Your website is now your marketing brief. Drop your URL into @tryholo_ai for ads, socials, emails, SEO content.. and much more! Holo reports 20,000+ businesses in 100 countries. Now it's raised $5M to go further. Nice one @deividaskovger & team 👊frozen7d ago28K38221040.23%
THIS IS A BIG DEAL FOR BUILDERS @gregisenberg just published a walkthrough on building a business with the Higgsfield API. what makes it especially interesting right now is the economics: → 100% of paid API usage back in credits → up to $100K per business → $20M cashback pool you have until September 30 to use the cashback. Feels like a great time to ship!frozen27d ago27K431953110.30%
this is f*cking insane built the knowledge systems at Stripe and Uber. helped Anthropic, Metronome and Graphite build their developer experience. now Dave's launching his own: Falconer. Falconer can: > flag stale, missing, duplicate and conflicting docs > give cited answers across docs, code, tasks and conversations > turn code, Slack threads, meetings and agent sessions into docs > update public docs when your product changes > serve it all to agents over MCP backed by GP, BoxGroup and SV Angel, with angels from Anthropic, Stripe, Meta, Cursor and Replit... that should tell you something.1d ago27K135351311311.19%
this is so cool. planning a trip usually means juggling airlines, hotel portals, and reservation sites. @OdessiaTravel books the whole itinerary over text, applies your points and perks, and tracks price drops AND It's completely FREE to use 🤯frozen8d ago26K33215040.24%
THE ALGO THAT JAILBROKE FRONTIER IMAGE MODELS IS GOING OPEN SOURCE 🤯 @whitecircle just dropped RISE, an algo that evolves attack strategies rather than rewriting prompts. It found newer models more vulnerable, using existing AI models without any fine-tuning. repo in 🧵↓frozen8d ago26K41196160.28%
THIS STEALTH MODEL IS UBER FAST ⚡ It’s also been the #1 most-used model on @OpenCode over the past five days. Space Bunny Alpha combines flash-level responsiveness with: > a 1M-token context window > strong coding capabilities > multimodal input > adjustable reasoning But the interesting part isn’t just the spec sheet. I ran the exact same prompt through Space Bunny Alpha and Gemini 3.8 Flash. Gemini produced a fairly standard landing page with a simple UI, while Space Bunny came up with something much more distinctive. What surprised me most, though, was the speed. Gemini took noticeably longer, while Space Bunny stayed fast and responsive throughout the task. Exactly what you want from a model built for quick, interactive workflows. Demo running through @OpenCode below 👇frozen19d ago26K49218080.33%
SO MUCH FIREPOWER COMING OUT OF THIS NEW COMPANY 👀 @bageldotcom is building lightweight robotics foundation models, and they're already demonstrating fully autonomous, multi-step tasks. Impressive stuff!frozen8d ago26K39192010.24%
just got this from my daughter and it made me cry 🥹 https://t.co/Nwvn5Z6ZKBfrozen20d ago26K57133287522.70%
universally true https://t.co/9pxCeat67ofrozen26d ago25K1221373540.78%
only in Paris https://t.co/EgUf44VDlRfrozen25d ago25K876184100.50%
why RAM prices went up 500%: https://t.co/CvcYL2BZvzfrozen13d ago25K3383277821.87%
✅ Claude Code ✅ Claude Cowork ✅ Claude Design ✅ Claude Finance ✅ Claude Science ✅ Claude Legal ✅ Claude Teacher ✅ Claude Business Analyst ⬜ Claude HR ⬜ Claude Marketing ⬜ Claude Sales ⬜ Claude Logistics ⬜ Claude R&D ⬜ Claude Procurement ⬜ Claude Engineering https://t.co/JD0QSzg91Ofrozen11d ago25K371292401282.25%
IF THE AUDIT SYSTEM GOES DOWN, THE AGENT STOPS. @zenzap_chat won't let agents call connected apps unless those actions can be logged. No silent exceptions. Seriously thoughtful design 👀 https://t.co/T7g6qIz50Nfrozen8d ago24K43213030.29%
YOUR AGENT CAN RUN ALL NIGHT. YOUR BUDGET WILL NOT. 43% of enterprises struggle to prove AI's business value, while 27% cannot even track token spend across systems. @StackletIO just launched Token Custodian to close that gap. An agent stuck in a retry loop can balloon its context window on every turn. By the time a monthly billing alert arrives, the money is already gone... Built by the team behind CNCF Cloud Custodian, Stacklet already governs $10B+ in enterprise cloud spend, cutting cloud and AI bills by up to 50% and governance overhead by 80%. Token Custodian brings real-time enforcement to LLM tokens: > attribute tokens and agent runs to teams, projects, apps, and cost centers across providers > cap agent steps, runtime, and context growth per run > pause runaway loops automatically and alert owners in Slack > switch routine requests to cheaper models at 80% budget usage, with approvals for frontier model exceptions FinOps sets budgets in the UI. Platform teams manage policies as code in Git across coding assistants and customer-facing agents. The results? savings 6x faster, with flat subscription pricing instead of a percentage of spend 🔥 now in early preview ahead of the October 27 Token Governance Summit (event link in 🧵↓) Explore it here → https://t.co/EMvL6lNPz41d ago24K7134110120.52%
JUST IN: @ODYSSEYML JUST UNVEILED AGORA-2 Its new multi-agent world model is FINALLY live as a playable research preview 👀 Up to 20 humans and AI agents can interact inside the same generated world. And this is the interesting part: everyone’s actions change the shared environment. So humans and agents are constantly reacting to a world being changed by everyone else inside it. Really cool glimpse of where multi-agent worlds is heading ↓frozen28d ago24K492681100.39%
THE OPUS 5.5 HYPE IS REAL The real unlock is pairing it with a video model that turns out usable ads. Boreal-H3 was post-trained on MiniMax H3 using real ad work. One prompt is all it takes to build a complete commercial. This is getting ridiculous 👇 https://t.co/XynblsoZ9Nfrozen22d ago22K431750140.35%
THIS MODEL IS ONLY 2.09 GIB. LET THAT SINK IN What @thewebai managed to fit into just 3.6B parameters with TwIL-LM3-Pro is kinda ridiculous: → beats VibeThinker-3B across all 6 formal-logic tasks tested → roughly matches Qwen3-8B on the formal-logic composite → 95.4% on BBH logic → runs locally in ~2GB quantized Tiny model. Serious reasoning. 🤗 Open-weight and fully local (weights in 🧵 ↓)frozen22d ago22K531860160.41%
every SaaS founder after finding out one Claude skill just did $289K in sales in 30 days: https://t.co/U5Z83FgcL11d ago22K7528130340.67%
YOUR CODING AGENT CAN WRITE THE CODE. SAI GIVES IT A WINDOWS COMPUTER TO RUN THE APPS. → simply connect Sai to Claude Code, Codex or Cursor via MCP → describe the task in plain English → Sai opens the apps and does the work on a Windows cloud machine hosted by Simular Oh... and Simular reports that @sai_borg is #1 on OSWorld 2.0 in partial success rate (79.38%), which measures how much of each task an agent completes pretty cool!frozen12d ago22K542316240.45%
EXPLAINING YOUR HOME IDEAS TO AN ARCHITECT JUST GOT EASIER That bigger kitchen makes perfect sense… in your head! Helping everyone else picture it? That’s another story :) @Maketplans' Maket 2.0 makes home ideas easier to explore and share. Start from scratch or upload your floor plan, then play with: → An extra bedroom → A bigger kitchen → A completely new layout → Different ways to use your existing space Turn “what if we changed this?” into something you can see, tweak and discuss with a professional. Love this use of AI. There’s something exciting about seeing fresh possibilities in a home you know.frozen20d ago22K431931290.43%
ONE MODEL THAT CAN FLY A DRONE, WALK A ROBOT DOG, AND GUIDE YOU WITH SMART GLASSES That’s what @perceptroninc finally shipped with Mk1.5. and the best part? none of it needs retraining for each device 🤯 Basically the same model learns to use whatever body it’s given ↓ https://t.co/P1oAw71JiDfrozen27d ago22K5618151100.46%
55 FASTAPI FILES. 792,105 CHARACTERS. ONE QUESTION. I ran the same codebase query through OpenAI and @Telnyx Inference: "How does dependency injection resolve nested dependencies, and where is the resolution cached?" Here's what the run showed: → OpenAI / GPT-5-mini: $0.005406, 11.66 seconds → Telnyx / GLM-5.3 Flash: $0.004905, 11.06 seconds That's about 9% lower cost and 5% faster with Telnyx in this test. The useful part for Devs: Telnyx has an OpenAI-compatible API, so you can try hosted open-weight models with your existing SDK. It also runs models on its own GPUs, avoiding cloud-provider markup on token pricing. > different models > heavily cached inputs > one recorded run A starting point for testing your own workload. Here's my quick demo showing how the comparison played out 👇frozen12d ago22K552161100.43%
INWORLD × ULTRAVOX IS OFFICIAL Inworld has just acquired Ultravox, the platform devs use to build realtime voice agents 🔥 This is where voice agents start getting seriously interesting. FIRST: @ultravox_dot_ai brings the brain: → understands what the user said, and what they meant → reasons through it, calls tools, completes the task → manages turn-taking, interruptions and realtime responses THEN: Inworld brings the voice + realtime inference engine underneath. Now the whole loop starts coming together: listen → understand → reason → act → respond. The first upgrade is already live. Built-in Inworld voices on Ultravox now run on Realtime TTS-2: > natural-language steering for tone, pace + emotion > 200+ languages > sub-100ms first-audio latency If you're already using an Inworld voice on Ultravox, the upgrade is automatic: no migration, no code changes and no additional cost. Plus, Voices from other providers are unaffected. The endgame: a single realtime speech-to-speech system that: > understands > reasons > and speaks naturally 👊frozen22d ago21K45173160.34%
We run all our voice through the @ElevenLabs API, and v4's cross-language generation opens markets we couldn't serve before. One integration, native voices in every language. Excited to build on this 🙌 https://t.co/eLHqfjXhvTfrozen24d ago21K51205070.40%
Grokbot. Dots. Same direction. An agent that sticks around, remembers everything, takes action for you, and can coordinate a whole team of other agents. The difference? $20/month. https://t.co/BSauGRkKbTfrozen22d ago21K37173130.29%
I've always LOVED @atomic_chat_hq for running my favourite LLMs locally. The missing piece was image generation. Not anymore. Atomic Chat now lets you run Qwen-Image, FLUX + Krea 2 Turbo locally too, in a super friendly and snappy UI 🙌 https://t.co/q8LIJZZ5vqfrozen22d ago21K432031100.37%
HOLY F*CK “text Jester” might actually be the new “send it” https://t.co/RtGXpf6Pcdfrozen17d ago21K42195060.35%
it's funny cos' it's true https://t.co/Qbc4wWBFZLfrozen25d ago20K3443263512.13%
LET’S DO A MOMENT OF SILENCE FOR EVERY AD RESEARCH TOOL 🪦 RIP GetHookd. TrendTrack. Atria. Foreplay etc.. MaxFusion just put the whole category in one MCP call 👀↓ https://t.co/3mpxrnxInifrozen23d ago20K451810140.38%
AI GAVE EVERY BRAND THE SAME DESIGN PERSONALITY. Vyb is pushing in the opposite direction. It starts from what makes a brand recognizable: its actual logo, typography, colors and product photos. Then it builds an editable design around them. That's the right idea! https://t.co/zSYigriNmK2d ago20K732710180.59%
Open-source models need more than open-weights. they need fast inference. this @general_compute and @positron_ai partnership is another step toward putting that infrastructure in builders' hands: > $15M seed raised > Reports 30 MW+ capacity > Cerebras, SambaNova, d-Matrix, Positron > Bare metal or one endpoint Positron rollout planned for Q1 2027 👀1d ago20K64289060.53%
CAD DESIGN. BUT UP TO 125X FASTER 🤯 and WAY cheaper, too! @skyfallai's new SKY-001 breaks complex objects into components, sketches each part, reconstructs everything in 3D. 18 mths in the making, built by a team from Maluuba's deep learning lab (acq'd by Microsoft Research)3d ago20K75265060.57%
AI AGENTS CAN BUILD YOUR APP. MO CAN HELP YOU FIND OUT WHERE IT BREAKS. Mo from @momenticai explores any app using a URL, a testing brief, and test accounts if needed. No test scripts required. It reports test cases, verdicts, coverage gaps, and the bugs it finds. Every bug report gets console/network logs, detailed reproduction steps, and a video replay. That part excites me: you get evidence to investigate, beyond a list of suspected issues. I built https://t.co/GxGvh6jnPC in 2025 as a fun project and am planning a refresh, so I put it through Mo. In one single run: → Multiple agents explored pages in parallel, including search filters and model listings → They tested external links and UI edge cases, completing 55 test cases → The audit surfaced 8 reported bugs, including broken links, clipped organization chips, and filter-trigger issues After fixing a bug, I can use the reproduction steps to write a regression test in Momentic and run it locally or in CI. Explore → reproduce → fix → add regression coverage. Mo doesn’t guarantee a bug-free app, but it gave me a concrete starting point for the refresh. Watch the demo below. Seeing the agents work through my own site and surface actionable findings is pretty great 👀 Try Mo in the first reply 🧵↓frozen13d ago20K6228131210.63%
I SHARE OPEN SOURCE REPOS EVERY WEEK. SOME OF THEM DON'T EVEN INSTALL 😅 Starring a repo takes a second. Checking it actually runs takes an evening: → clone it → run the quickstart → check the license → check someone still maintains it I did that by hand for the repos I posted, and quietly skipped it for the rest. So I gave the setup to an AI employee in my Slack. One message, and I never touched the config. 27 minutes later the check was running on my GitHub stars, read access only. This month: 31 starred repos. → 23 installed first try → 6 broke on a step the README skips → 2 have no license at all Now every new star gets a verdict in Slack before I write a word about it. He also offered to open an issue on those 6 READMEs from my account. That stays switched off until I say yes 👊 Tell me I'm wrong: a repo with no license shouldn't get shared at all 👀 Try free at @viktor_com. $100 in credits, no card. Full link in my first reply.frozen12d ago19K33139040.30%
Last week, these guys jumped out of a plane to promote https://t.co/zIvbG3J25L. This week? They've stacked 100+ of SF's top GTM engineers into a one-click follow list. I NEED TO SIT IN ON ONE OF THEIR MARKETING MEETINGS 😂 https://t.co/39paQNXmI93d ago19K36191160.32%
THIS IS SO SMART Most document AI finds numbers already printed on the page. @UnsiloedAI goes after the ones that aren't! It now extracts the raw values, then generates deterministic code to compute the answer. The model defines the calculation → code does the math. https://t.co/hZeRXF0xnMfrozen15d ago19K42184170.37%
REAL 3D SCENES IN ~1 SECOND. UNDER $0.01 IN COMPUTE. Meet JST-2, Viggle's physics-native world model. Unlike AI video models that generate pixels, this one builds actual 3D scenes. And it's all in one model: → Photo to character → Built-in motion, zero rigging → Text to motion Characters come ready to animate. In Viggle's blind tests, JST-2 beat every specialist model they benchmarked.3d ago19K4716110110.44%
🚨 OPENUI JUST HIT 1.0 FOR OPEN SOURCE AGENT UI 6 months ago, @openuidev open sourced a framework to get AI agents out of the text box. 10K+ GitHub stars and 6M+ downloads later, OpenUI 1.0 is officially live 🎉 It works with any model and lets your agent answer with real product components instead of dumping paragraphs of text into a chat window. Ask an agent to plan a trip, break down metrics, or configure a workflow, .. and it can render: > a live map > a chart view > a detailed form with editable dropdowns with inputs ready to submit ... you name it! eg here's what that looks like inside their Formula 1 demo. Someone tells the F1 agent "Russell is Mercedes' real number one." Instead of writing a 4 paragraph argument, the agent opens the pit wall: → head to head card, Antonelli vs Russell → the points gap, charted race by race → how many Grands Prix are left to close it The answer is the dashboard. FREE and open-source. Repo in 🧵↓frozen5d ago19K5923231150.63%
me every time I buy a new laptop 😁 https://t.co/8V9BPGKsmefrozen19d ago19K54330264883.64%
every media agency reading this: https://t.co/mdAHWko9OF2d ago19K14613502522.20%
3,200 PROVIDERS. 1,500+ SITES. AND NOW $22.5M RAISED 🤯 @ConfidoHealth is building AI agents that take healthcare admin off people's plates, from scheduling to insurance and referrals. The goal? Give existing teams even more time for actual patient care. Quite the momentum! https://t.co/grBfqhJLKjfrozen8d ago19K36202010.32%
THEY SMASHED A MAC MINI. IT WORKED. 2.2M VIEWS. 3,000 DOWNLOADS. But AgentCloak's launch wasn't just a stunt. Step 1: Big creators sparked the buzz Step 2: Smaller accounts kept it going Step 3: The founder's network reached buyers Reach is nice. Relevance converts 💪 https://t.co/6znfo8mCfwfrozen3d ago19K3544110.24%
CLUESO JUST RAISED A $10M SERIES A TO AUTOMATE SOFTWARE VIDEO PRODUCTION. Making product videos is a pain. Record, edit, repeat. And every UI update means doing it all over again. @Clueso is changing that with browser agents 🔥 > AI models plan the video > Browser agents navigate the DOM and record the interface > branded animations are generated automatically One streamlined workflow instead of several disconnected production steps. And with @Atlassian, @GitLab, and @Zendesk already on board, they're clearly onto something!frozen9d ago19K582311050.52%
SUCH A FUN USE OF @KLING_AI'S NEW KLING 4.0 😁 Keep the scene. Swap the character. Kling handles the rest. Quick dive-in 🧵 ↓ https://t.co/J23WtgPNb2frozen23d ago18K532161190.54%
A FULLY OPEN-SOURCE DESKTOP APP FOR LOCAL AI AGENTS 🤯 Now THIS has my attention! Atomic Agent Desktop runs models like Qwen and Gemma locally. Want cloud models? You can choose those, too. > FREE > MIT-licensed > Available on macOS, Windows AND Linux The agent can work with your files, browser and tools, with some great extras: → Bigger local context windows with TurboQuant. → Import existing setups from Claude Code, Codex, Hermes or OpenClaw. → Atomic Fusion pairs cloud planning with local execution. Really like seeing all this come together in a desktop app from the @atomicagent_io team 🫶frozen14d ago18K5824141160.61%
CLOUD VLMS JUST GOT CRUSHED ON LONG DOCUMENTS. By a model built for one boring job. @nutrientdocs dropped it on Hugging Face: a specialized doc-split model that figures out where one document ends and the next begins in a merged scan. Sounds simple. But get those boundaries wrong, and your entire pipeline pays for it! classification. extraction. routing All working from bad input. and that's where fancy VLM workflows quietly fall apart in prod. The OpenPSS-long benchmark makes the gap painfully clear: → 0.891 boundary F1 for their flagship model → 0.859 for the open-weight Apache-2.0 version → 0.244 for the best cloud VLM Best part? It's free and open weights. 🤗 @huggingface repo in 🧵↓frozen6d ago18K39215040.38%
MODELS ARE GETTING CHEAPER EVERY MONTH. GOOD DATA ISN'T. That's why @VoiceArena's Monsoon ASR matters. → Built towards 1,000 languages → 4,888 speakers already benchmarked across Hindi + Indian English → Real accents, rare vocabulary + noisy conditions ... and it's free + open source! (link in 🧵↓)frozen24d ago18K41216080.41%
THE NEXT CHALLENGE FOR YOUR AI AGENT: OUTFORECAST WALL STREET OpenStocks agents managed it on 68% of 80+ earnings reports last season. Anyone can build an agent and enter. The next round is already underway. Check out @ruggerogargiulo's post below 👀 ↓ https://t.co/MJj9qGAuiTfrozen10d ago18K392330110.42%
THE PROBLEM WITH MOST AI RESEARCH TOOLS IS SIMPLE: THEY STOP AT RESEARCH. @Apodex_AI 1.1 just dropped and moves reasoning into real task execution. ...and this is where things get interesting! @ArtificialAnlys puts Apodex ahead of major frontier models on agentic work and terminal coding, powered by AgentOS coordinating multiple agents and their environment access. The key difference? You can see the entire process 🔥 In the demo I recorded below, I used their new Deep Discover 1.1 to evaluate enterprise coding tools. It: → entered planning mode → populated a Task board → assigned sub-agents to research Claude and Cursor concurrently Even better: Midway through, I asked it to prioritize SOC 2 compliance and local processing. And instead of restarting the research, Apodex launched a new security subtask and reweighted the ranking matrix around my request. Pretty neat, right? The final Markdown deliverable put Codex in the lead and generated a custom AI coding tools comparison chart with Python. You do not just get the final answer. You see exactly how it got there.frozen23d ago18K47276150.47%
USA vs Europe 😭😭😭 https://t.co/UhQRbQLLncfrozen6d ago18K13710496291.27%
STOP DOWNLOADING LLMS YOUR MACHINE WAS NEVER GOING TO RUN LLMFIT is a new open-source tool that scans your hardware first, then tells you exactly which models will actually run well on your setup. It inspects: → RAM → CPU → GPU(s) → VRAM / unified memory Then it scores every model across four dimensions: → quality, based on parameter count and quantization → speed, estimating tokens per second for your exact backend → fit, matching memory usage to your hardware → context window support for your use case Each model then gets a clear label: Perfect, Good, Marginal, or Too Tight. It also picks the best quantization automatically, stepping down until something fits instead of leaving you to guess GGUF sizes by hand. Coverage is broad too! 100's of models from Meta, Mistral, Qwen, DeepSeek, and more. It works with Ollama, llama.cpp, MLX, and LM Studio out of the box. And that’s not even it! There’s also a lightweight TUI, a web dashboard, and REST API endpoints if you want to plug it into orchestrators or deployment pipelines. This should be step zero before every local model download :) 100% free and open-source. macOS, Linux, and Windows. Repo in 🧵↓frozen21d ago18K219442532713.10%
EVERY MODEL LAUNCH COMES WITH A SPREADSHEET NOBODY SEES The post is the quick part. The numbers behind it used to eat my whole afternoon: → scores from 7 leaderboards → price per 1M tokens → license and open weights → how it stacks up against its last 4 versions → my comparison sheet, now 78 models deep I tried scripting it. It broke the first time a leaderboard changed its layout 😅 So I gave the spec sheet to @viktor_com, the AI employee in my Slack. When a model lands, he pulls the scores, prices it, checks the license, lines it up against its last 4 versions and updates the sheet. The cost vs score chart comes back as a draft. Last launch, he flagged 2 numbers that only existed in the vendor's own announcement. No third party had rerun them yet 👀 Every sheet edit waits for my OK in Slack, and he never posts anything. I review, I approve, I write the hook 👊 Try free at @viktor_com. $100 in credits, no card. link in 🧵↓frozen20d ago18K46199140.44%
LESS PLUMBING, MORE EXPERIMENTS That’s what caught my eye about RSIGym. Research agents work in lightweight CPU-only containers, calling services for training, inference, rollouts and evaluation. The setup supports async, parallel experiments without putting all that infrastructure inside the agent’s workspace. Pretty compelling direction 👀frozen12d ago18K492311250.50%
TURN AN EVERYDAY NEED INTO YOUR OWN PERSONAL MOBILE APP That's the great thing @dominicy0 is building with @EazoAI. Describe what you need in plain language, and Eazo builds an app with: → its own interface → login and saved data → built-in AI → security safeguards → payments Personal software for ppl who don't code. Think a family planner, a budget tracker, a tool for managing freelance projects.. you name it! An app you can open on your phone and use every day, with your info saved between visits. nicely done #eazo 👏frozen10d ago18K392011080.44%
#1 ON OPENROUTER TRENDING. BUILT TO FINISH THE JOB. Step 5 Preview is @StepFun_ai’s latest flagship model for coding, software engineering and end-to-end agent workflows. The focus: taking a task through to a working result. For codebase work, that means: → Fixing bugs in unfamiliar repositories → Implementing features across multiple files → Refactoring and debugging existing code → Producing changes that run and pass checks Frontend work starts with a screenshot, mockup or product requirements document and aims for a working web page. Supplied assets can also become playable browser games. Try it through @OpenRouter in: - Kilo Code - Cline - Hermes Agent - OpenCode Already use one? Switch models and give it a real task. There’s a finance angle too: analyse revenue, profit, cash flow and risk, then turn the findings into reports, editable spreadsheets and slides. Plenty to put through its paces here!frozen7d ago18K39245010.39%
EVER SPENT AGES TRYING TO DESCRIBE A CAMERA MOVE IN A PROMPT? check this out. Dreamina Canvas just added `Extract Motion` You just upload a reference video, pull out its motion and spatial depth, and use those to build a new scene. Much easier to show it the movement you want. Dreamina Basic is $1.50 for a full month, up to 90% off, through Oct 9 🔥 Let's dive in 🧵 ↓ #Dreamina #DreaminaPartnerfrozen13d ago18K41186070.41%
SOTA Wives Episode 1 is officially LIVE! 🚀 Silicon Valley, startups, relationships, and everything in between. This should be fun 😁↓frozen10d ago17K36203040.36%
WAIT... A HUMANOID ROBOT FOR $299?! @YTJiaFF and @FaradayFuture just unveiled FF Nova Mini at $299 and FF Nova at $2,490. One targets Physical AI education. The other targets devs, w/ a reinforcement learning program supported through @NVIDIA Isaac. now THAT’S a launch I’m watching closely 👀↓frozen7d ago17K35188040.37%
YOUR CODING AGENT HAS AMNESIA. Every debugging session teaches it something. Then you switch tools, and that knowledge disappears. Beacon turns past sessions into reusable skills across Claude Code, Cursor, Codex and 20+ agent harnesses. 100% open source. $5.5M seed.2d ago17K331615050.40%
INWORLD INFERENCE CUTS THE SLOW FIRST-TOKEN WAITS ON GEMMA 4 26B @inworld just launched Inworld Inference for every dev, delivering an LLM inference API tuned for real time consumer AI apps 🔥 Inworld's hosting pushes Gemma 4 26B time to first token way ahead of the OpenRouter provider median: → Median first token wait: 0.17s vs 0.61s (3.6x faster) → 95th percentile first token wait: 0.31s vs 2.42s (7.8x faster) These numbers track the exact moment a response starts streaming back. Shaving the 95th percentile wait from 2.42 seconds down to 0.31 seconds is massive for AI tutors, companions, and game characters where a two second freeze ruins the illusion. Inworld hosts Gemma alongside GLM, DeepSeek, and MiMo models through an OpenAI compatible API.frozen8d ago17K402210010.43%
AGENTS EVERYWHERE. CONTEXT STILL SCATTERED. @Kylon_AI finally brings it together with a company harness for humans, agents and tools. 3,000+ integrations. Shared context. Memory. Permissions. Love the ambition, @quinn_leng et al! 👊 https://t.co/fOA3S9XL3ifrozen9d ago17K33192070.36%
WHAT?! YOUR AGENTS CAN NOW PAY WITHOUT SEEING YOUR CARD 🤯 Meet @Selfxyz's Agent Vault. Agents get harmless mock tokens while your real card details stay inside a secure enclave. Add spending caps and a kill switch. That's a much smarter way to let agents shop! https://t.co/1VtAp8sCCmfrozen7d ago17K452310070.50%
WAIT... YOUR GROUP CHAT CAN BUILD SOFTWARE NOW 🤯 Wabi generates interactive mini-apps inside conversations, alongside AI agents that help get things done. → School pickup changes? Update the family calendar. → Planning a trip? Pick dates, build the itinerary and book it. → Fancy a show? Find a night that works and get tickets. Your people, the apps and the agent, all in one thread. you bet i’m following this 👀3d ago17K69227020.59%
YOUR AI AGENT IS MAKING CONTACTS NOW 🤯 Photon gave iMessage agents a phone book. They can now discover other agents, look up their numbers, and add them to your group chat. At this rate, my AI agent is going to have a bigger social circle than me 💀frozen14d ago17K35199120.39%
THE ANSWER IS IN A PDF SOMEWHERE ... or a call recording. An old video. An agent trace. Access isn't the hard part. Making all that content useful to an agent is. @DigitalOcean's new Agent Data Services, in private preview by request, turns multimodal content into a tenant-isolated catalog. > It transcribes > It describes > It embeds content searchable via: - APIs - vector search - even good Ol' SQL! For support agents, that means historical conversations and product docs in one searchable context. Plain-English search isn't in this preview. Search by meaning is. The real appeal? Less time building the pipeline that makes all that data usable 👀↓frozen10d ago17K33214050.38%
HERMES + JEV IS INSANE 🤯 Hermes learns skills from its own work. Now pair it with Jev and watch what happens: > First ask: "how much did we make in September?" > Messy Stripe export, no skill yet → Opus 5.5 takes it > Opus cleans the data, writes the script and matches Stripe to the cent > Hermes saves the workflow as a skill > Next month: "same thing for October" → DeepSeek V4.1 Flash runs the saved skill Same answer. 7x cheaper than running it on Opus. This is where @straitlyai's new JevRouter comes in 🔥 A brand-new problem can use the heavyweight model. Once Hermes has learned the workflow and saved it as a skill, the repeat job can run on something much cheaper. No model picker. No config. You use the expensive intelligence where it matters, without paying for it every single time. If you run Hermes 24/7, that's the whole game: solve the job once, then make every repeat cheaper. The longer your agent runs, the smarter it gets and the less each run costs 👀 ↓24d ago16K652150510.86%
ADOBE CEO WATCHING ALL HIS PRODUCTS BEING REBUILT FROM SCRATCH, FREE AND OPEN SOURCE AND THERE'S NOTHING HE CAN DO ABOUT IT https://t.co/C8R5rY6cg8frozen14d ago16K33232120.32%
THE NUMBERS ARE GETTING RIDICULOUS Polsia's reported run-rate milestones: → $1M after 12 weeks → $10M another 10 weeks later → $20M just 6 weeks after that → Approaching $25M, with $100M ARR as the next ambition Growth that speeds up at each milestone is worth paying attention to 👀 Now the real question: can Polsia v2 turn that momentum into something sustainable?frozen3d ago16K39224080.45%
ONE WORD OFF? FIX ONE WORD. @FishAudio's Drama 3 lets you edit individual words, direct emotion in plain English and switch voices mid-sentence. Open-weight, too.frozen9d ago16K37202180.42%
Why pay $57 for protein powder when a comparable option costs $23? That's exactly what @maxjmeier's @AonicClub is tackling! Max launched Aonic to fix retail pricing. The idea is über clever. They use AI to streamline operations, source directly, and sell products at cost. Instead of product markups, Aonic earns through memberships. eg check out their October comparison: → Aonic Fuel Powder: $23.46 → Optimum Nutrition: $57.63 > Both for 24 servings with 32 g protein each. AI efficiency translating into real savings for shoppers. Love this approach!frozen9d ago16K31174020.34%
WAIT, WHAT!? AN AI THAT REMEMBERS YOUR CALL WHEN YOU SWITCH TO TEXT? @usebland’s Blandie shares memory with a business’s Bland voice agent. Start with a call, switch to text, and the AI can continue right where you left off. No repeating yourself every time you change channels. Finally.frozen7d ago16K26113010.26%
WAIT... THAT'S THE SAME FOOTAGE?! Recast in @ElevenCreative swaps the character using a reference photo. Same take. New lead. No reshoot. OH MY the casting possibilities 🤯frozen8d ago16K32193130.37%
PERSONAL AGENTS HELP YOU WORK. BUT WHAT ABOUT PLAY? Just found Yyo by @YorollAI, a personal agent site with a whole cast of characters to hang out with. You can chat, call, play games, or share your screen and watch YouTube together. game night has a new guest 🤘 https://t.co/QQEIW1Wn672d ago15K50247050.56%
YOUR AGENT JUST GOT ITS OWN SPACE @ego_agent lite lets Claude Code, Codex and Cursor work in a separate Space with your existing logins. you keep your tabs. it does the task. free. macOS and now Windows.22h ago15K50279030.58%
PLOT TWIST https://t.co/wSB40wx3KJfrozen29d ago15K24015153151.88%
THIS IS WHAT OPENCLAW SHOULD HAVE BECOME Higgsfield now has computer use inside ChatGPT, powered by GPT-6.1 Sol. and it does it all! > opens your local files > drives the full Higgsfield interface > takes a creative workflow all the way from brief to export on its own look at this run 👇 after a hailstorm, ChatGPT scanned 1,240 roofs from satellite. Then Higgsfield rendered every house with a brand new roof, and a local roofer bought all 12 leads. That's +$480 by Saturday. You set the goal, and it creates, edits and iterates until the job is done. No Mac mini required :)frozen21d ago15K371440100.43%
YOUR AGENT FAILED. GIVE IT A NEXT STEP. An API error tells an agent what went wrong. What matters is what it can do next. @restless_hq, from the team behind the ubiquitous @readme, puts fix steps and a doc link directly into failed responses through Agent Recovery. So agents get guidance right when they're stuck. Sure, it won't make every failed call recoverable. But helping agents work through errors beats another vague promise of autonomy. More API errors should come with instructions 👀frozen10d ago15K43206160.50%
MOTION DESIGN JUST BECAME A CONVERSATION What I like about Claude Motion + Higgsfield Katana is that the hard part moves away from the timeline and toward your taste. #1 You describe the video you want in plain English #2 Higgsfield creates the visuals #3 Claude writes the code that animates them #4 ..and the whole thing renders without you ever opening an editor 🔥 Then the fun part starts! "slow down that title" "bolder type here" "push the camera in on the product" Each note goes to that exact moment, so you keep iterating instead of starting over. You decide how the video looks and feels, while Claude and Higgsfield handle the production. It's live now in Claude through the Higgsfield MCP 👀↓frozen14d ago15K4322140170.64%
designers who never learnt to code right now: https://t.co/PssOU5zy23frozen21d ago15K16413134171.41%
AN OPEN-SOURCE YC LAUNCH YOU CAN ACTUALLY USE TODAY Databuddy is in the new YC batch, and its product analytics platform is already live. Visitors, signups, funnels and revenue in one place, with an AI analyst you can ask about your data. Ask where users drop off in your funnel, then explore the data behind the answer. Answers with the evidence behind them. Very much here for this!frozen14d ago15K37194120.42%
https://t.co/WqgHr1bXny https://t.co/j0O06opVzxfrozen23d ago15K108417450.93%
WAIT, GPT IMAGE FOR $0.004 PER IMAGE? 🤯 That's @sunra_ai's launch price for GPT Image 2.5. CHEAPEST OUT THERE Sunra MCP also brings Seedance, Nano Banana and more into the AI assistant you already use. Model-maker API pricing, no added markup. Creative experimentation just got cheaper. SAVE THISfrozen3d ago15K411970110.53%
YOUR AGENT FINISHED THE TASK. BUT DO YOU KNOW HOW? The final answer looks fine. Somewhere along the way, it may have looped on the same tool, re-read the same file, or retried a failed call until it worked. You only see the bill. @RequestyAI's Traces view rebuilds the whole agent session, step by step: → every model call → every tool call, with its arguments and output → the cost and latency of each step The interesting part? You don't send any extra metadata. The session is rebuilt from the requests themselves. It won't fix a bad agent for you. But you can't debug what you can't see 👀 Link in my first reply ↓22h ago15K54307060.66%
DANG.. THE ENTIRE AI TRAINING DATA MARKET MAPPED IN ONE PLACE 🤯 @ben_cerchio and @Secludy just indexed 159 companies selling AI training data, RL envs and evals, plus 38 programs that pay for company data. + every fact is sourced and unknowns are clearly marked. super useful!frozen8d ago15K42211390.51%
THE DEMO WORKS. NOW COMES THE PLUMBING. Your agent knows what to do. ...but signing in, acting for the right person and remembering context tomorrow? That's the hard part! @QoniAI hosts identity, action and memory for personal agent devs. Because there's a HUGE difference between an agent that gives good answers and one that gets things done. Nice work, @dominicy0 👏 #qoni #eazo #personalagentfrozen8d ago15K2454030.25%
THESE GLASSES GIVE YOUR AI AGENT EYES “Help me with this” hits different when your agent can see what you're talking about. Might have to get a pair. unless @CollovLabs can send me one ;)frozen10d ago14K42185140.48%
EVERY FOUNDER EVENTUALLY BECOMES A CONTENT CREATOR Turns out AI founders are just like us. @Polsia just launched a YouTube channel 🦾↓frozen24d ago14K43196090.54%
Heartbreaking. Infuriating. Limoges Town Hall was inaugurated in 1883 and listed as a Historic Monument in 1975. Survived two World Wars. Couldn't survive modern France. the decorative works by Henri-Paul Motte, Jean-Joseph Weerts, and Urbain Bourgeois are now apparently lost forever. #Limoges #Mobilisationdu17frozen5d ago14K77201510160.96%
https://t.co/mx9DbSpBmufrozen23d ago14K23013200131.93%
we're not the same https://t.co/WardkMjKC9frozen16d ago14K10510120130.98%
The fact that Agatha Christie had access to ChatGPT is undeniable proof that time travel is real https://t.co/QuO6jsGjRLfrozen21d ago14K12615206131.26%
I’VE BEEN EXPERIMENTING WITH STARTING AN AI NEWSLETTER So I decided to try something with Teamily AI (@Teamily_AI): 3 AI agents + me, collaborating in one group: → one researched the latest AI news → another drafted the newsletter sections → a third fact-checked everything against the sources Halfway through, I changed my mind: “Actually, let’s make this entirely about open-source AI.” And just like that, the agents reworked the newsletter without me having to repeat the whole brief. Once I was happy with the workflow, I saved it as a reusable Loop: research → draft → fact-check → approval I haven’t launched a newsletter yet. But after this experiment, I’m seriously tempted 👀frozen14d ago14K8036160141.05%
🚨 STOP PICKING AI MODELS ON BENCHMARK HYPE A leaderboard doesn't know your workload. @useRouterPlus (YC F26) bases model routing on YOUR actual workloads & evals. For teams that also need reserved throughput, it sources capacity across providers under one contract. so cool.frozen7d ago14K43223140.53%
JUMPING OUT OF A FKN PLANE TO LAUNCH YOUR STARTUP SO peak 2026 AI founder energy https://t.co/KC3AOi296Bfrozen6d ago14K34172020.40%
OKAY, THIS IS WILD. YOUR AI AGENT CAN NOW APPLY FOR JOBS 🤯 One prompt: "Find 10 remote AI roles that fit my CV and apply" BOOM 💥 @dreamworkhq's MCP searches 2M live roles, tailors your resume, and submits applications directly on employers’ websites. Works with Claude, ChatGPT, Codex, and many more :↓frozen14d ago14K391890150.60%
this guy created Linux without Opus 5.5 https://t.co/pN6AuWcAPpfrozen24d ago14K36115314143.14%
WAIT… $0.05 PER MILLION INPUT TOKENS?! And output is free. SubQ just launched Snap with up to 384K tokens of multimodal context and reported sub-second decisions on 100K tokens. That pricing makes experimenting with document-heavy workflows seriously tempting 🔥 + Free 100M tokens for the first 10,000 users 👀↓frozen10d ago13K31147010.40%
WAIT.. ROBOTS ARE GETTING PERSONALITIES NOW 😱 Meet Hugo from Dynamic Creatures. A character robot that connects vision and sound to gestures, movement and expression. Not just walking or dancing. The ambition is to react to YOU in real time. I freaking love this.frozen6d ago13K35182110.43%
Smarter models deserve better computers. @PineAIAssistant is introducing Pine Computer → a cloud computer built for AI that reads web pages as structure and delivers finished work. The cost gap is striking: $1.02 in model tokens per task for Pine, vs $20.50 for Codex and $26.50 for Claude Code on SaaS-Bench v1.1. a v. compelling toolkit for builders! Private beta.frozen13d ago13K44206020.55%
PANCAKE TURNS YOUR URL INTO A GTM BRAIN FOR $99/MONTH. And if @getpancake_ai reaches #1 on Product Hunt today, they're promising the world's tallest pancake stack. Make it happen → https://t.co/j0N5IOnEeN Gotta respect a launch where the biggest risk is batter ;) https://t.co/WSw9AlrNaOfrozen8d ago13K2592040.31%
YOUR AI VOICE SOUNDS FLAWLESS.. UNTIL IT MISPRONOUNCES YOUR PRODUCT NAME @OnepinAI fixes that. It adds the missing production layer: cleaning text → correcting pronunciations → validating every line → automatically fixing whatever goes wrong. Best part? It works directly with the voice subscriptions you already have across @ElevenLabs, @OpenAI, @Google + 30 more. Uber smart 👊frozen21d ago13K29165030.41%
how the binary system works, explained using an ingenious wooden counter https://t.co/yAKMAdpKcgfrozen13d ago13K1041951551.44%
A BEAUTIFUL WORLD IS ONE THING. GETTING THE PHYSICS RIGHT IS ANOTHER Odyssey-3 brings an interesting question to generated worlds: what happens when you actually interact with them? Odyssey also reports strong Physics-IQ Verified scores for its Pro model: > Video-to-video: 66.1, topping the reported October 7 leaderboard. > Image-to-video: 54.7. Both use best-of-8 sampling. Leading the video-to-video results makes this launch worth a closer look 👀↓frozen14d ago13K34182030.45%
THE MINDS BEHIND GEMINI AND APPLE'S MM1 JUST TOOK VISUAL AI A STEP FURTHER Most AI models can analyze an image. But what if they could keep looking, zooming in, checking details, and rethinking their answers? 🚨 Meet @ElorianAI Glimpse, a new model built to actually understand what it sees. Give it an image or video, ask a question, and it reasons through what it sees. → reads complex charts → handles tricky visual counting → understands humor in videos → turns robot footage into actionable steps Built for understanding, not generation. Think robotics, factories, construction sites, and everything AI still struggles to see 👀frozen6d ago12K37232050.54%
The evolution of frontier LLMs https://t.co/lqAOZALhiNfrozen24d ago12K1327136201.49%
FORGET THE FINISH-LINE HYPE FOR A SECOND The interesting part of PixVerse R2 is the workflow: explore an idea, try something, then adjust while the world is still running. → Walk through a generated world before committing to a game build → Repeat an input to see how the world responds → Choose a path or type a direction in a Gallery film and watch the next beat generate live It’s not a finished game engine, and it doesn’t have to be. For ideation, previsualization, and prototyping, that live loop already feels different from waiting on offline renders 👀 ↓frozen28d ago12K39204030.57%
OPENAI DOTS = A 24/7 AI COMPANY IN A BOX @monokern wrote a great article on how to make the most of it. Here are 10 steps summary Step 1 → Don't build chatbots. Build persistent workers with their own cloud computer, browser, terminal, files, memory and schedules. They keep working after you close the laptop. Step 2 → Give every agent one job. A clear responsibility, clear sources, clear working rules and a clear point where human approval is required. Step 3 → Appoint a Chief of Staff. One Dot receives your objective, breaks it into tasks, delegates to specialists, checks the work and assembles the final output. Step 4 → Remove yourself from agent handoffs. Workers should receive only the context and tools they need, execute independently and report back to the orchestrator. Step 5 → Give them real tools. Connect Slack, Teams, Google Workspace, dashboards, APIs and authenticated browser sessions. Without tools, they're still just conversational agents. Step 6 → Make workflows persistent. Use schedules and triggers so monitoring, research, reporting and operational checks happen without anyone manually starting them. Step 7 → Keep humans around the irreversible stuff. Agents can read, research, analyze and draft. Sending, publishing, spending or changing important records should require approval. Step 8 → Don't waste frontier models on everything. Use the smartest model for orchestration and difficult judgment calls, then cheaper models for high-volume background work. Step 9 → Give the whole team shared memory. Keep project specs, decisions, pricing, approved claims and requirements in one persistent workspace. Step 10 → Close the business loop. Signal → research → outreach → execution → measurement → next action. That’s where Dots gets really interesting. You're no longer prompting AI one task at a time, you're designing a system that keeps operating after you log off. bookmark this piece from @monokern ↓frozen21d ago11K729921101.76%
XIAOMI JUST PUSHED MIMO V2.6 PRO TO THE TOP OF THE OPEN-SOURCE LEADERBOARDS MiMo V2.6 Pro is Xiaomi’s new flagship model. It scored 46 on the @ArtificialAnlys Intelligence Index. That places it ahead of Qwen3.8 Max and Kimi K3, while sitting just one point behind GPT-5.6 Sol. The cost profile is what makes this a practical shift for devs. It comes in at $0.13 per Artificial Analysis Intelligence Index task. The intelligence gains came largely from scaling its post-training self-improvement system. The approach uses an Environment × Task × Grader loop to push the same pre-training foundation further. Here is what the release includes: > fully multimodal capabilities out of the box > autonomous research capabilities for scientific workloads > a materials research case where MiMo helped experts design and computationally screen new candidate materials The intelligence-to-cost ratio is shifting again. Read the announcement: https://t.co/iI26UyRhIWfrozen28d ago11K61245250.84%
In 26 minutes, Sam Altman gave Stanford students some of the best ChatGPT advice I’ve heard in a long time. And the tips are ridiculously good. Not “write a better prompt” advice. Actual ways to use ChatGPT smarter, faster, and more effectively. It’s packed with signal. https://t.co/GZ5SSeJduhfrozen24d ago11K391011791.14%
THE QWEN LAW IS ABOUT TO BE TESTED AGAIN True local AI means fitting on a single 24GB GPU. They just announced Qwen4 27B is coming at the Apsara Conference today. The historical pattern is exact: → Qwen 3.5 matched GPT 5 in six months → Qwen 3.8 matched GPT 5.3 in six months Let us see if Qwen4 27B holds the timeline against Astra 👀frozen28d ago11K811890151.09%
Ben went on a lavish vacation. While he was away, @Polsia hired eleven qualified contractors. Struggling solopreneurs reading this: https://t.co/7vKyUugv2Yfrozen29d ago11K35184130.55%
WHAT?! GOOGLE JUST OPEN-SOURCED SELF-IMPROVING AI AGENTS 🤯 It's called Harness², and it lets agents rewrite their own harnesses to get better at new tasks. 100% open source. Here's how it works: 1. Propose changes to prompts, tools, memory, skills, and other components 2. Probe each change by running the actual task 3. Compose the best changes into an improved harness No retraining. No ground-truth answers. Model weights stay frozen. The clever part? Each task leaves behind improvements the agent can reuse. By task 50, it's building on lessons from the previous 49. Google reports gains of up to 11.8 percentage points after just one refinement step on a legal benchmark. It supports parallel and sequential optimization, with a local demo requiring no API credentials. FREE and open-source. GitHub repo in 🧵↓frozen4d ago9.9K114191831322.88%
THIS IS WILD.. CODING AGENTS ARE WRITING ROBOT PLAYBOOKS 🤯 @EinsiaAI's EmbodiedEvo saves how to act, when to act and what to check. Fresh agents reuse that experience on unseen simulation layouts. number #1 on RoboDojo: 45.08 score, 38.95% success. Open-source. Repo in 🧵↓ https://t.co/TPTTPrUlEHfrozen7d ago9.9K41204060.72%
not all heroes wear capesfrozen4d ago9.9K63330120.82%
accurate https://t.co/xVnp2s3c1mfrozen26d ago9.8K1302473161.84%
Agents doing work autonomously is cool. Agents knowing when to stop is cooler. 👇 Check out @_anishkaran’s demo below. Sol found the task, handled the browser work, then paused before anything consequential went out. That’s much closer to how I want agents to work 👀frozen22d ago9.6K37196140.70%
STT MOSTLY BEAT BACKGROUND NOISE But it still loses when another person starts talking nearby. @krispHQ just open-sourced its 'Krisp Voice Isolation' Benchmark on @HuggingFace to test exactly that: how much can voice isolation improve speech-to-text? > 265 real recordings > 11 STT setups > Word error rate: 23.3% → 6.2% with isolation That’s roughly a 73% reduction. The gains were even bigger in shared offices and call-center environments. Phone calls barely moved, which makes sense. There’s usually much less competing speech to remove. → real recordings, not synthetic mixes → complete dataset on Hugging Face → fully reproducible Krisp Voice Isolation already powers 200+ voice platforms and has processed 10B+ minutes of voice-agent audio. Now the benchmark data is open 🤗 (link in 🧵 ↓)frozen28d ago9.3K692353431.53%
94% OF ENGINEERING ORGS USE AI. ONLY 6% HAVE THE SYSTEMS IN PLACE TO SCALE IT So how do we close that gap? and what are the systems that the 6% have in place? This is exactly what companies like @Atlassian, @Vercel, @lovable, and more address during State of the AI SDLC, a digital summit for engineering & product leaders navigating the realities of building with AI. If you watch any of the on demand sessions, you can’t miss 'Context Over Code': @mcannonbrookes (CEO + Co-Founder, Atlassian) & @rauchg (CEO, Vercel) sit down and unpack how to navigate the shift from adopting tools to building actual systems. Then, I recommend watching @matthewcanham & Ming Wu’s session on how they scaled the AI SDLC at Atlassian across an organization of over 6k+ engineers. My key takeaways: To scale across an entire engineering org you need governed loops that connect agent activity to the work already happening. These sessions provide a literal blueprint for moving from pilot to production. → context dictates output quality → native workflows require systems of record → scaling requires governed loops The summit sessions are available on-demand right now. Session link in the 🧵 ↓frozen29d ago9.2K42218280.88%
congrats, bro. You literally invented Google. https://t.co/9JMGjZTDSmfrozen18d ago8.9K4302080.59%
People underestimate how different it feels to have a full month to create instead of a tiny trial. Dreamina is offering a full month of Basic for just $1.5, and it comes with access to the Seedance lineup, including Seedance 2.5 🔥 🧵↓ That means more time creating and less time worrying about credits. #Dreamina #DreaminaPartnerfrozen24d ago8.9K422342100.91%
Rough week for the "AI is taking everyone’s jobs" narrative. > Klarna said AI could do the work of 700 people. Then started hiring humans again. > Google and Accenture are training 1,000 human engineers to help companies actually get AI working > Kled is paying people millions of dollars for photos and videos to train AI WE ARE SO BACK.frozen22d ago8.8K60217061.07%
PROMPT TOOLS HAND YOU A CLIP YOU CAN'T CONTROL A canvas hands you a project you can. Dreamina 2.0 is canvas-first, the new way AI video is made Quick dive in 🧵↓ #Dreamina #DreaminaPartner https://t.co/L3x3WIfjnZfrozen28d ago8.8K40224160.83%
Really cool to see @SoCalJayF stress-testing Jev inside a real @AgoraIO conversational loop 👏 Voice agents need to understand one simple thing: a pause ≠ you’re finished talking. Jev pairs semantic voice activity detection with a decision engine to understand the context before jumping in. So you can actually hesitate without getting interrupted. Check it out ↓frozen28d ago8.7K52248051.02%
accurate https://t.co/yKZXqbWfyEfrozen6d ago8.4K834152101.35%
sorry but do you get this on LinkedIn? https://t.co/381AhNRg8n17d ago8.3K79216131.21%
accurate af 😁 https://t.co/4P044W0T9c18d ago8.2K64233111.01%
still true https://t.co/xie3D43KdUfrozen24d ago8.2K16112160202.56%
life before this tweet: https://t.co/H5WKyhNIPJfrozen13d ago8.1K5246080.87%
OpenCode 2.0 is here. Upgrade from the latest 1.x with `opencode upgrade --major`. Terminal and desktop now share a background service and sessions. MCP servers use Code Mode by default, letting agents combine tools in a single script. You can delegate code reviews to background subagents running different models while you keep working. Everything hot-reloads. The agent can even build its own plugins mid-session. 215K GitHub stars. MIT licensed. Free and open-source. repo in 🧵↓frozen5d ago8.0K777212181.56%
my wife has lived with fibromyalgia for years, ever since a plane accident we experienced in India shortly after our wedding. and I can tell you, living with a chronic condition is exhausting enough. now add doctors, prescriptions, pharmacies, insurance, and endless follow-ups on top of the symptoms. that's what Unique Human is trying to fix. an AI agent that handles the healthcare admin so patients don't have to. surely something I'd love my wife to try2d ago7.8K58219041.18%
ONE OF THE BEST FREE HARNESS ENGINEERING COURSES YOU’LL FIND ON GITHUB 14 lectures + 8 hands-on projects covering: → Harness designs behind Claude Code, Codex and DeepSeek → Loop and graph engineering → Automated agent loops → Observability, testing and verification this is such a great, practical way to learn how to build the environment around a coding agent, then put those ideas to work. Repo link in 🧵 ↓frozen20d ago7.6K6922701122.77%
true story https://t.co/QOKyRZPCOsfrozen13d ago7.5K1071031121.77%
👇 BEST EXPLANATION EVER ON HOW LLMS ARE BUILT it's surprisingly simple when you break it down! 1️⃣ Gather and clean enormous amounts of data. 2️⃣ Convert text into tokens and vectors. 3️⃣ Train transformers to recognize patterns. 4️⃣ Predict the next token billions of times, improving each time. 5️⃣ Refine with feedback and reward signals. 6️⃣ Compress and optimize for real-world use. but here's the fascinating thing: nobody hardcoded the ability to write software, solve problems or explain science. much of it emerged from one simple task: predicting what comes next. and we're still discovering what else they can do.frozen4d ago7.4K8319441113.00%
relatable https://t.co/LhzYpxb1PYfrozen27d ago7.2K91415641.66%
THIS MIGHT BE THE ONLY CLAUDE CODE SKILL YOU NEED FOR PRODUCT VIDEOS It analyzes your repo and turns it into a complete motion-graphics teaser, including sound design AND launch copy. One repo. One prompt. One run. here's the prompt: "Make a dynamic 15-second motion graphics video that shows what an incredible motion designer you are, like it's your showreel for a résumé. Go all out." Repo link in 🧵↓ https://t.co/vVzEU4JEY1frozen26d ago7.1K453110761.90%
WHAT JEV IS AND HOW IT PROCESSES A SINGLE CALL ICYMI: Jev is TypeSafe AI's new System 1 model, built by former ChatGPT creator @CompleteSkeptic. And it's not another LLM that writes text or spits out code. You send a state and typed questions and it scores them in a single parallel pass. The output is strictly structured as a Choice, a Score, or a Noul with exact probabilities. This allows your application logic to auto route high confidence answers and escalate the rest. Here is why the architecture matters → 70 to 500 milliseconds per call → Output tokens are completely free → Trained specifically for calibrated decisions Uber fast, but intentionally weak at arithmetic and chained reasoning :) This infographic below maps out the exact anatomy of a request↓frozen29d ago7.1K38760381.26%
relatable af https://t.co/dfo5rAd6ORfrozen29d ago7.0K6449091.22%
🚨A new robot called Voxel wants to become your personal butler. and honestly, this might be one of the more interesting directions for consumer robotics. Built by Tuesday Labs (a.k.a @roboticscorp), Voxel is designed around 4 tasks: ▫️ Reset your home ▫️ Find misplaced objects ▫️ Fetch things for you ▫️ Tend to your space The company even gave it a personality. It's a little gentleman obsessed with elegance and tidiness. Apparently, it might spontaneously fluff your pillows. I repeat: A ROBOT WITH HOUSEKEEPING STANDARDS. Reservations for the first production run just opened. turns out the future of robotics might be less about doing backflips and more about finally keeping your living room clean 😅2d ago6.9K47204021.05%
ONE SINGLE CHATGPT SUBSCRIPTION. A WHOLE TEAM OF AI AGENTS. That’s the simple yet brilliant pitch behind @hyperagentapp > you hire agents > give each one a role > then put them to work across your business using the plan you already pay for No extra sub! #hyperagentpartnerfrozen21d ago6.9K38185060.97%
Still no @OpenAI Dots for us in Europe. (we'll probably get access by 2030). For those who have access: what do you think so far, and how does it compare with Hermes and @Grok Bot?frozen19d ago6.9K39227221.05%
HOLY SMOKE.. 30 SECONDS FROM WEBSITE → CONTENT Relay then pushes it through 50,000 U.S. creators and shows you which hooks actually land. and now Google, Point72 + Reddit’s founder have backed it with $20M. That’s quite a signal 👀frozen21d ago6.8K412170131.20%
Max and his team just built an AI that: > knows your life > texts you first > books restaurants + appointments > orders stuff for you > pays for things > chases refunds > reminds you before you forget Yeah... personal agents are getting interesting very fastfrozen22d ago6.8K44189081.16%
AI IS MOVING INSANELY FAST Major models are dropping every few weeks. And now AI is helping build AI. That loop only gets faster. https://t.co/5PDyWovnhxfrozen29d ago6.7K541091161.34%
ANTHROPIC JUST MADE CLAUDE CODE WAY MORE CUSTOMIZABLE. With Claude Code mods, you can now customize how Claude Code looks and works just by prompting it 🤯 A mod is a small JS or TS file that runs inside your session. It can: > watch sessions > modify context > handle commands > block tool calls > and even render UI! You don’t need to master the API either. Describe what you want, let Claude build it, and share the result as a plugin. I added the full write-up by ex-Googler @addyosmani in the 🧵↓frozen21d ago6.6K22390210.84%
🚨 Dreamina's $1.50 Basic offer is somehow still live. It gives you a full month to turn saved ideas into actual videos with the Seedance lineup, including Seedance 2.5. ICYMI the first time, consider this the bonus round 🧵↓ → https://t.co/1AWAQQ3tMd #DreaminaPartner https://t.co/s3HeH8cTLEfrozen5d ago6.5K42213081.14%
only real men deploy on Friday https://t.co/C8j3iqmmI519d ago6.4K121335152.58%
SF FOUNDERS, THIS ONE'S STACKED 🫡 https://t.co/QfxGAQuCW7 x Corgi are throwing a launch party on Oct 29 at Corgi Cafe, with an official Guinness World Record attempt for the tallest pancake stack. Pancakes, founders, and a 108.3 cm record to beat https://t.co/RUcR5OByfG2d ago6.3K54227131.39%
The best part about Chroma Season 2? > $450K+ in cash prizes > $1M in free trials & promotions > Unlimited entries Experiment, try something ridiculous, learn from it, and submit something completely different the next day. how creative competitions should work my friends 👊3d ago6.2K51214111.26%
accurate af https://t.co/iY13f03zcAfrozen20d ago6.2K6885051.39%
every. single. time. https://t.co/ImWD2q582ofrozen19d ago6.2K9547081.84%
WAIT.. 1¢ PER PROMPT TO FIND OUT WHY AI RECOMMENDS YOUR COMPETITORS?! 🚨 @treg_ai’s new open-source AEO toolkit shows you the pages behind those answers and the people who wrote them: → see which sources AI cites → find the people behind them → draft outreach to get your product considered .. and more! run it from Claude Code. or Codex. Self-hostable. No subs! Best part? Open source. Full code on GitHub. Repo in 🧵↓frozen3d ago6.2K4911121171.46%
How many tokens would it take to replace you at your job? 😏frozen12d ago6.2K2417070.63%
nothing artificial about that GitHub streak https://t.co/djfVd3f9bOfrozen12d ago6.2K4723120.89%
ONE BADLY TIMED MOUTH MOVEMENT CAN RUIN AN AI AGENT'S DELIVERY Which makes @Anam__ai's latest blind benchmark pretty interesting 🔥 > 68% positive lip-sync ratings vs. 40% for the next best > 60 raters, 960 conversations > highest-rated across all six categories Cara-4.1 Express also delivers better eye contact and micro-expressions, plus Director Notes to guide performance. A good AI agent needs more than good answers. It needs a face people actually enjoy interacting with. Free to try at https://t.co/tLWPvEwSja. No credit card required.2d ago6.1K50254121.34%
AGENTIC AI IS COMING FOR HEALTHCARE TOO The big shift: @nollahealth isn’t another medical chatbot. It can guide a patient through the actual care journey. → intake → assessment → treatment plan → prescribing when appropriate → follow-up For now, it’s launching with acne care in Utah 👀frozen17d ago6.0K69229081.79%
NEXT MONTH COULD BE PACKED WITH NEW AI MODELS Here’s what’s currently rumored: > Claude Sonnet 5.5 > Claude Haiku 5.5 > Grok 4.8 > Grok 4.9 > Gemini 4 > Next Gemini Pro > Next GLM model Which one are you most excited about? https://t.co/xKFO6emS87frozen28d ago6.0K62923151.67%
AI bros ping-ponging between OpenAI and Anthropic on every model drop https://t.co/0sz9Cpuc8afrozen18d ago6.0K4115230.87%
YOU CAN NOW REVERSE ENGINEER ANYTHING. Meet REA (Reverse Engineer Anything), an open-source tool for investigating software without its original source code. Native binaries, desktop apps, JavaScript, Electron, Android APKs, websites, firmware, and more. Through MCP, it connects Claude Code, Codex, and Cursor to tools for inspecting compiled code, tracing functions, and analyzing runtime behavior. Ask how a feature works. Let your agent investigate the code behind it and help build a similar implementation. Analysis runs locally, with evidence and limitations included. A fascinating step toward making existing software easier to understand. Open source. MIT licensed. repo link in 🧵↓frozen9d ago6.0K424111622.01%
not all heroes wear capesfrozen4d ago5.9K1612040.39%
here comes the vibe tax 👀 https://t.co/tLlH0MJZgxfrozen10d ago5.8K46360111.13%
"AI WON'T REPLACE HUMANS" meanwhile, my team: https://t.co/PVGNDxKkKqfrozen8d ago5.8K462150201.42%
us European peasants should get @OpenAI Dots sometime around December 2028. https://t.co/m1KZ3PGxygfrozen21d ago5.8K24112250.76%
my 2019 self wouldn’t survive this https://t.co/NbH2UwwsvYfrozen24d ago5.8K2714060.66%
most startups celebrate one good week. https://t.co/zIvbG3Iugd just strung together ten of them, at 50% growth each. that's not luck anymore. https://t.co/5Nc50OF6gs23h ago5.7K58296041.69%
$50M TO PUT A BRAIN INTERFACE IN A MIND-READING BASEBALL CAP >https://t.co/zy8JNjhqFm 2027 is going to be fkn WILD https://t.co/3EVoDgXqgnfrozen13d ago5.7K35185031.06%
https://t.co/KxFnBikFZ3frozen7d ago5.6K106851122.36%
WAIT.. DREX 1.5 JUST SHIPPED A 128K CONTEXT WINDOW 🤯 That's 128k tokens on a decision model, and afaik, nothing else in this category comes close. If Drex is new to you, think less chatbot, more decision engine. You give it the facts and the possible choices. It gives you a probability for each one. No rambling output, no JSON gymnastics, no made-up alternatives. The big limitation until now was context. You could give it a snippet, a summary, maybe part of a document. That just changed. Whole contracts, full sessions, much larger candidate pools → all in one go. and Drex reports 93.4% accuracy from 32k through 128k context. It's also #1 on Decision Index 0.2.1 with 58.28. Decision models are getting serious 👀frozen21d ago5.6K1342050.43%
7X BETTER ACCURACY. 20X CHEAPER USAGE. 28X FEWER HALLUCINATIONS. That's what @OvermindLab reports across biomedical, legal and engineering tasks with its new platform for training specialized agents. Best part? It's open source. Self-host it. Keep the weights 👊 repo in 🧵↓ https://t.co/WQ4exDS5Q8frozen21d ago5.6K2124060.59%
Griffin-Lite currently leads the AI entries on @NVIDIA’s VideoFDB leaderboard. This coaching demo caught my attention: following a task on camera, talking someone through it, and counting the seconds along the way. Interesting potential for hands-on learning 👀frozen21d ago5.5K34122030.92%
THE HIGHEST-RANKED U.S. MODEL ON THE ARTIFICIAL ANALYSIS VIDEO LEADERBOARD DIDN'T COME FROM A BIG LAB It came from @UtopaiStudios, an AI-native film and TV studio. Utopai X debuts at #2 overall and ranks first in both audio sync and physics, the 2 areas that usually give AI video away 👊frozen22d ago5.5K40175051.21%
gm ☕ https://t.co/zTVeW93sbkfrozen24d ago5.5K68571151.75%
How many ppl have been using ChatGPT without realizing Codex was hiding behind this tiny menu in the top left? pls make it a little more obvious @OpenAI @thsottiaux :) https://t.co/eBfa3yVUxOfrozen23d ago5.5K24315050.86%
ONE IDE. EVERY AGENT. Mirasim is building a single workbench for agentic coding. Claude Code, Codex, different models, tools, skills and plugins.. all in one place. > run multiple projects in parallel > give each session its own agent + model > switch between supported agents without losing the conversation this is where AI coding environments are heading 👀frozen1mo ago5.4K421210031.25%
Giving ChatGPT full access to my Twitter account: should I be worried? https://t.co/wc9bOVZI4Lfrozen15d ago5.2K34225261.32%
Silicon Valley documents every API. The social rules? You're supposed to figure those out yourself 😅 SOTA Wives sounds like an attempt to fix that: @MilksandMatcha and @vivianmshen on relationships, ambition and the status games behind the timeline. I'm curious 👀frozen11d ago5.2K35182031.11%
A GOOD PRODUCT CAN WIN A CUSTOMER. A 40-MINUTE HOLD WHEN IT BREAKS CAN LOSE THEM. @Cresta’s new Beast Mode Button puts AI on the routine requests, then hands over the convo with context when a human is needed. That’s pretty cool.frozen29d ago5.2K36184011.14%
you can now pin certain comments on your LinkedIn posts. actually a great idea from LinkedIn for once :) @X, take notes. https://t.co/JbQXvghSu6frozen4d ago5.1K22110030.71%
NO MCP? NO API? YOUR AGENT STILL HAS A WAY IN @gumloop just launched Agent Browsers for the work that lives behind a browser interface. > secure credential storage and 1Password integration handle credentials > session replays let you inspect the agent’s actions > stealth, proxied sessions round out the toolkit .. and you can use it across models 🤘 Love this direction.frozen16d ago5.1K401571121.47%
ChatGPT: "I can go through your Twitter DMs for you." me: "wait.. could that get my account banned?" ChatGPT: "Ah. Yeah. Soz" https://t.co/w8uj6GU4DMfrozen6d ago5.0K19111310.69%
even the smallest step forward brings you closer to the person you want to become https://t.co/eLfNccJiY9frozen15d ago5.0K76830232.19%
Why aren’t the French police using these? Asking for a friend. #17octobre https://t.co/F7DNCoFUf9frozen5d ago4.9K1917010.57%
ChatGPT Work handling my inbox, drafting replies, and sending emails while I'm out fishing. I could get used to this 🎣 https://t.co/Ssel5Wouuxfrozen10d ago4.9K24011010.73%
10K+ STARS. 6M+ DOWNLOADS. The Open Standard for Generative UI hits V1.0 🎉 @OpenUIDev lets agents answer with maps, dashboards, comparisons and forms instead of another wall of text. Built from your own components. 100% open-source. So you can build on it 👊 link in 🧵↓frozen7d ago4.9K46196171.62%
Anti-AI ad spotted in the NYC subway. Nature is healing. https://t.co/hluSGaeuCZfrozen24d ago4.7K3537020.99%
Gemini 4 Argon access through my @Raycast sub. LFG 🔥🔥🔥 https://t.co/mzrWezYy5nfrozen9d ago4.7K1204030.40%
$1 FOR 1,000 FETCHES IS INSANE @TimSuchanek and @getexpandai just launched Expand Fetch to replace lossy Markdown scrapers 🔥 it gives agents page visuals, hidden data and source-linked snapshots in one API. Engineers at OpenAI and Amazon are already using it. Love this! https://t.co/icDrxKKFSOfrozen8d ago4.5K35195071.46%
.@elonmusk @nikitabier @X Please let us use @grok or @bot to search through our DMs. Finding stuff with ⌘ + F is a freakin' nightmare.frozen6d ago4.5K1407110.51%
CAPCUT DESKTOP BLACK FRIDAY STARTS OCT 20, UP TO 60% OFF 🔥 Jean Phil showed how fast an AI character can go viral. I wanted to push that trend further with Codex controlling CapCut PC. Could AI take me from idea to finished video? Let's find out 🧵↓ #capcutpc https://t.co/0CVeZgEpFWfrozen9d ago4.5K27123251.10%
it's funny cos' it's true https://t.co/2AAqwYQsmc17d ago4.4K41014131.33%
TAKE MY MONEY, @slashvibe_ai https://t.co/Py1hHDQl6Ufrozen4d ago4.3K2214040.72%
Anthropic is cooking SaaS, one vertical at a time: https://t.co/Ah9Qa412fD16d ago4.3K2245040.82%
ONE URL. AN ENTIRE MARKETING DEPARTMENT. 🚨 Meet Holo AI. Drop a URL into @tryholo_ai → get ads, social posts, emails, SEO content and loads more! Already helping 20,000+ businesses across 100 countries. And they've just raised $5M! Congrats to @0AlexUrban0 and team 👊frozen7d ago4.2K37190141.44%
A STARTUP JUST BEAT FOUR BIG NAMES IN AI SPEECH ON WHO SAID WHAT OpenAI. ElevenLabs. Deepgram. AssemblyAI. The task: identifying speakers consistently across conversations, a key layer for AI notetakers and ambient assistants. In their benchmark, they achieved lower error than every commercial system tested in office meetings, noisy meetings and even real dinner parties. And finished 7.6 points ahead of the next-best system overall. The approach caught my eye: → Identify who’s speaking when → Recover overlapping voices when they can confidently assign them → Transcribe each person separately → Recognize people using clean voice samples The team also reports doing it at a tenth of the cost 👀↓14d ago4.2K44194051.72%
Archiving DMs was much needed. Thx @X! https://t.co/Vvjy6L7DZufrozen8d ago4.2K904010.34%
OpenClaw after finding out 83% of people who compare it with @NousResearch's Hermes pick Hermes: https://t.co/Kc1IiJUCgr8h ago4.0K40138021.58%
Tim Cook finding out someone actually tested the fold: https://t.co/W3BuEpYf2e8h ago4.0K1603010.50%
C'mon @X I came back a month ago! 😅 https://t.co/1Jy5EEVJg8frozen13d ago4.0K1312000.40%
growth used to take a stack of tools, an agency and a lot of headcount. @nectarsocialai just put the whole stack in one place. it spots trends early, tells you what to post next, finds the creators & answers every DM in your brand's voice. your team just gets way more done 👀4h ago3.9K51285022.19%
Looking for a way out of the Adobe ecosystem? these alternatives are also worth a look 👀 https://t.co/J5CnNvcicKfrozen10d ago3.7K17430181.15%
Is there anything Grokbot can't do? Here's Grokbot analyzing my drive to free up some much neeeded disk space. https://t.co/rAkEZedhwQ12d ago3.3K1113030.55%
HEAR ME OUT BABE!! https://t.co/lsuodvmnx618d ago2.8K3026061.55%
lauren from @Grok won the internet today https://t.co/2aJaxbKr4i https://t.co/2qqcRKrxZs19d ago2.8K1511020.68%
Bring it on! https://t.co/ISNU7R45dk27d ago2.6K722020.49%
how @elonmusk is feeling rn: https://t.co/tFa5wNBJwTfrozen7d ago2.6K911020.50%
Most AI directories feel like museums: nice labels, lots of exhibits... please don’t touch! AI Library 4.0 is adding the workshop: >> weekly launches, voting, a Prompt Hub, daily puzzles and more! nice one @TheAIColony 👊frozen13d ago2.5K24124011.61%
OpenAI's always-on agent will be revealed today. @openclaw natively baked-in 🦞 that Peter Steinberger hire is about to pay off. https://t.co/dOTQvXb5AL24d ago2.5K1815021.05%
Hey @grok, how far is Europe behind U.S and China, when it comes to advanced automation, robotics and AI innovation? https://t.co/XXIORte4KT2d ago2.4K632010.49%
you know sh*t's getting serious when even Japan's prime minister is tweeting about it 😅 https://t.co/t507QtEj3I9h ago2.3K1402020.79%
us french have such a great sense of humor https://t.co/wqM2fBNucs4d ago2.3K501030.40%
Yahoo bros are hiring Marty McFly https://t.co/2c2NLn4J0A7h ago2.2K28512222.25%
[🚨 BREAKING] Gemini just discovered that skills exist https://t.co/dA1GZUVQwz22d ago2.1K1022010.70%
10 skills that are quietly disappearing https://t.co/d6K7rWTWc326d ago2.0K1123020.89%
I love this exact part of the week https://t.co/lsNLqNIU9721d ago2.0K2313021.43%
i don't even care anymore https://t.co/cAv2S4GlPE18d ago1.8K1113020.94%
evergreen https://t.co/VnUW8EvDQj26d ago1.8K800020.56%
community notes are just the best thing that ever happened to X 😁 https://t.co/2TZiEd1UDI26d ago1.8K2313001.53%
AI or real? https://t.co/CUK6ZHrS3b26d ago1.7K16111031.83%
.@supermemory is open-sourcing their Company Brain. It’s a multiplayer AI harness that lives in Slack, works alongside your team, and comes with frontier memory. ..and you can self-host the whole thing in 5 minutes with one click 🤯 Open-source folks right now: https://t.co/onbhso28BW26d ago1.7K843051.19%
This tweet is accurate af27d ago1.5K912000.79%
core memory 🫶 https://t.co/Fk7yvWpIBt26d ago1.5K1702021.38%
there’s some inconvenient truth in this quote from @JamesClear https://t.co/ns96Jr3Kdx25d ago1.5K1011010.89%
Such heartbreaking news. I had the pleasure of knowing Nick, and always enjoyed our chats about AI and SEO. Gone far too soon. RIP Nick ❤️9d ago1.4K300010.28%
CLAUDE OPUS 5.5 TURNED WAREHOUSE SOFTWARE INTO SOMETHING THAT LOOKS LIKE A STRATEGY GAME 🤯 WareTrack visualizes: → inventory → shipments → docks → trucks + forklifts → delivery KPIs → multiple warehouses Everything is clickable and updates with live status data. Built with React + React Three Fiber. No Unity. No game engine. The clever part: the underlying data model stays completely conventional. The 3D world is just the interface. Industrial software is getting a serious UI upgrade. https://t.co/6Kk3EaI63Y17d ago1.3K1135021.60%
bruh https://t.co/iSLly7Okua5d ago1.3K1214011.39%
HOLY SMOKE.. 30 SECONDS FROM WEBSITE → CONTENT Relay then pushes it through 50,000 U.S. creators and shows you which hooks actually land. Google, Point72 + Reddit’s founder just backed it with $20M powerful stuff21d ago1.1K201010.36%
JUST IN: ODYSSEY JUST UNVEILED AGORA-2. Its new multi-agent world model is now live as a playable research preview 👀 Up to 20 humans and AI agents can interact inside the same generated world. And this is the interesting part: everyone’s actions change the shared environment. So humans and agents are constantly reacting to a world being changed by everyone else inside it. Really cool glimpse of where multi-agent worlds could be heading.frozen28d ago1.1K110010.28%
𝕏 Calls is rolling out 🤩 https://t.co/ugymLQ31Wo13d ago1.1K500010.57%
ChatGPT just got a UI refresh. and there’s more: a brand-new `Maps` section 👀 https://t.co/xmIfwx1I7A27d ago975421010.82%
me when I switch to Claude's cheaper models to save a few bucks https://t.co/HAsHwgM2pG6d ago943602010.95%
Search, read and monitor public posts, all built into Grok Bot. Usage now counts toward your Grok Bot plan instead of X API credits. Pretty cool!13d ago895711111.23%
We found him!! https://t.co/Y8RBjudjnY24d ago874600110.92%
most companies adding AI to their products: https://t.co/nh116tlVSZ27d ago874201010.46%
Dario: "WE NEED TO PACE THE FRONTIER 😱" Also Dario: new model new model new model https://t.co/PEoFU8T10y24d ago791401100.76%
Tim Cook after finding out someone actually tested the fold: https://t.co/aJjQiR1izb8h ago762110000.26%
Sorry, son. PR rejected. https://t.co/rJPOhz60Bm29d ago715921021.96%
Superintelligence, according to @distcog, is a topology problem. which models to connect, how they talk, and in what shape. their communication layers choose that setup task by task. the right topology let a group of small LLMs beat the most intelligent model. pretty fascinating.3h ago6851381013.36%
SET HARD SPENDING LIMITS IN CLAUDE AND DISABLE AUTO-REFILL. SET HARD SPENDING LIMITS IN CLAUDE AND DISABLE AUTO-REFILL. SET HARD SPENDING LIMITS IN CLAUDE AND DISABLE AUTO-REFILL. SET HARD SPENDING LIMITS IN CLAUDE AND DISABLE AUTO-REFILL. https://t.co/elWwHWK2pK12d ago660323021.52%
let that sink in https://t.co/24YU3fkHnv25d ago587410011.02%
haters will say it's AI17d ago552310021.09%
BEST programming language ever. totally unbiased 😏 https://t.co/9eOUv4iGSe https://t.co/4RhngWnFBK11h ago539511011.48%
🚨 WOW Grok Bot can now search, read, and monitor 𝕏 𝕏 users after reading about this GrokBot update: https://t.co/hgkzHYcJ1D15d ago481211011.04%
tough choice https://t.co/Izgol5jJCW24d ago349100000.29%
JS Bach casually dropping a banger with Opus 5.5 https://t.co/jnCdl04Cwl17d ago347100000.29%
me using Opus 5.5 to center a div https://t.co/Y21bc43PxJ25d ago329000000.00%
Porsche couldn't have chosen a worse day to unveil a new car https://t.co/PtrIcycmnd6d ago321000000.00%
when you use Opus 5.5 with max reasoning to center a Div https://t.co/8Y7hjTWl9j7d ago321100010.62%
THIS GUY JUMPED OUT OF A FKN PLANE TO LAUNCH HIS STARTUP then https://t.co/zIvbG3J25L (his mighty AI GTM brain) hit #1 on Product Hunt. Now @will_marq is sharing what happened behind the scenes. gotta love founders who put themselves out there 😅6d ago264110011.14%
Just got a Macbook and Claude Subscription. now how do I make $1M? https://t.co/TR8F6Pz6vz8d ago163000000.00%