← Back to leaderboard
adxtyahq

aditya

Verifiedactive
@adxtyahq

Engineer

RosterAI Model News & ReleasesStartup & SaaS Builders
30,399 followers · 2,546 following · Last polled 3h ago
Perf. Score
50.7
Rank #230
Value Score
39.9
#167 of 273 priced
Median views
3.4K
p25 2.3K · trailing 7d
Followers
30K
+0.2% 7d
ER (impr.)
1.64%
eng ÷ impressions
ER (foll.)
0.54%
eng ÷ followers
Est. CPM
$59.24
QT-rate basis
WoW views
-27.7%
▼ falling

Rates & value · est. from 3.4K median organic views

overpriced
Quote tweetbasis
$200
$59.24 / 1K views
Post
—
no CPM
Thread
—
no CPM
Retweet
—
no CPM
Value Score 39.9 · rank #167 of 273 priced$2739.73 / 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 (3.4K), not the mean (9.9K) — so a single viral post doesn't inflate a creator's reach. The p25 floor (2.3K) is the level 75% of posts clear.

Recent posts · ranked by views

PostPostedViewsLikesRepostsRepliesQuotesBookmarksER
switched from MacBook M2 to M5 Pro today all my passwords, logins, extensions, apps, files, settings, literally EVERYTHING transferred over with one click no one is coming close to the Apple ecosystem anytime soonfrozen14d ago430K9.9K246237464922.54%
Preferably the best time to enter the AI space and get a job too. Cant say much but make sure to study langchain, langgraph, RAG, vector DBs, embeddings, reranking, LLM inference, quantization, evals, agents, MCP, vLLM, fine-tuning, distillation and GPU basics.28d ago402K6.4K47362186.9K3.44%
World’s most advanced model just launched. https://t.co/JviCqc6uBP
frozen
9d ago
280K
726
15
14
1
186
0.34%
my 16GB MacBook Air genuinely sucks in 2026 😭 I can’t even run Claude Code in the terminal alongside Docker, Codex, Chrome (Spotify + Slack) without everything starting to freeze. this exact setup was smooth last year. I still have ~70GB of storage free too. is this the apps getting more bloated or is macOS just getting worse?frozen22d ago113K1.1K527091471.35%
BRO THAT’S LITERALLY 3 OPUS 5.5 PROMPTSfrozen7d ago87K1.3K992821.57%
yesterday I was talking about how we STILL don't have AI agents that can shop for us based on our preferences. TODAY, I asked Hermes on Telegram to find the cheapest New Balance 550 in my size across 3 sneaker stores A few minutes later it came back with prices, stock and links, and told me to wait: none had my size, and Superkicks is cheapest if it restocks @Tiny_Fish's Web Agent did the browsing. @NousResearch's Hermes handled the rest we're getting closer to a future where i just tell AI what i want and it figures out the rest.frozen9d ago83K6735060.10%
THIS AIN’T TIER-69 BRO https://t.co/RutaDG98gLfrozen4d ago81K2.2K12220752.90%
switched back to chrome after using brave and safari for weeks google chrome is literally the definition of too big to fail.frozen12d ago79K2.4K4517916953.43%
why is Meta even asking “implement multi-head attention forward pass” in a Research Scientist interview 😭 https://t.co/6Y5PFUBBHffrozen19d ago77K652242949852.20%
Reverse engineering in real life https://t.co/jhWniwk0RIfrozen10d ago75K2.0K423331372.93%
with a founder profile like this ex-Apple, Amazon, Flipkart, London Business School, IIM Ahmedabad i don't care where my money goes, i'm all in https://t.co/9XJ7CuMaL12d ago71K358131331400.74%
bro these guys are fucking insane 💀 THEY'RE ACTUALLY TRYING TO SET A GUINNESS WORLD RECORD FOR THE TALLEST PANCAKE STACK 😭 https://t.co/N7vdUOmAnQ2d ago66K7239130.13%
earning in USD and spending in INR is genuinely one of the biggest financial advantages you can have in your 20sfrozen15d ago47K3.6K892221638.20%
one thing that’s helped me land opportunities is cold emailing YC founders i’ve been using this for 7-8 months, emailed 200+ founders and got replies from ~40 finally open sourcing the tool i used to do it pick a batch → get founders + emails → write once → send to everyone small tool, saved me a ton of time.frozen20d ago45K802274851.4K4.97%
Didn’t reply thinking it was some kinda scam, just to find out everything was legit… https://t.co/Et5wu1mkMdfrozen28d ago43K40023211081.27%
OpenAI is absolutely cooking this DevDay - 1. GPT-6.1 Sol at $2/M input + $10/M output with an ultrafast option 2. $500 Pro plan with no limits & 25x the usage of Plus 3. Decisions API that’s almost instant https://t.co/oQu4U0K5pnfrozen23d ago39K5424250.17%
If you just want to survive and not enjoy life, then definitely move out at “₹1.5-2 lakh” salary But if you at least want to have some fun, think about moving after ₹5L+ (and not US, Canada, Switzerland or Singapore)frozen20d ago39K4264931281.48%
companies like apple, google, microsoft and even accenture are paying above-market rates in india so they can scale faster they can cut costs significantly compared to hiring in places like SF, NYC or other major tech hubs while still getting engineers with the skill sets they need we already knew software engineering was booming in india, but seeing companies pay numbers like ₹1.1L/month for internships makes it even more obvious.26d ago37K48012113721.58%
How are newer image models are actually easier to jailbreak than their previous versions? according to this research, GPT-Image-2.5-Flare was ~3x easier to jailbreak than GPT-Image-2, while Nano Banana 2.1 was ~2x easier than Nano Banana 2 the fact that newer doesn’t automatically mean safer is a pretty big deal, especially when these models are already being shipped to millions of users🫠frozen8d ago34K9954050.33%
BRO WHAT THE FUK IS THISSS https://t.co/E5RzseOppjfrozen23d ago34K152571380.59%
bro went skydiving to announce a GTM startup 😭 the future of marketing might just be founders risking their lives for a product launchfrozen9d ago34K5808060.21%
what happens to software engineering in 2030? AI is already in every company, every team, even government software. where does that leave software engineers?frozen19d ago32K2199395470.99%
Learning Applied AI alongside hardware and low-level systems today is like learning to code in 2006.28d ago26K1.0K581306206.46%
if you’re looking for a job, try this: type “find 100 startups hiring software engineers, identify the right person to contact at each one and research the company” go to sleep and wake up to 100 companies, the right people to reach out to and all the context you needfrozen22d ago24K22051323252.38%
thats why i rented a 64-core, 256GB RAM instance i can now run 10+ chats simultaneously, spawn 25-30 subagents, keep docker running and have separate worktrees for everything without any lag or heat only con is it costs almost $800/month but its worth it if you love your device and value efficiencyfrozen11d ago22K1961393911.53%
NO FUCKING WAY 😭 these guys jumped out of a PLANE to launch their AI startup, hit #1 on Product Hunt, and now they're trying to break the GUINNESS WORLD RECORD for the tallest pancake stack 💀 HOW DO YOU EVEN TOP THIS 😭frozen7d ago20K2824040.19%
“Sundar is a monster” - actually makes sense now https://t.co/Ad18tJr6U8frozen22d ago18K229621231.45%
invest the next 3 years in AI, AI agents, multimodal AI, AI infra, and applied AI this is honestly the best time to get into the ecosystem, build things, learn fast and shape up your future around it.frozen21d ago18K799521843396.87%
This man has 12 degrees from Harvard, Stanford, Oxford, Yale, Princeton and more.... I think top college admissions is basically the first version of building a professional portfolio. You have grades, research, projects, competitions, essays, internships and everything else, but doing 20 random things isn't necessarily better than doing 3 that actually make sense together. I actually didn't realize there were companies helping students figure this out. And then Jamie Beaton, who apparently took this whole education thing very seriously 💀frozen27d ago18K715142390.75%
I was part of cohort-2 and back then it was actually fine because we didn’t have all these advanced AI tools that can teach you almost anything from scratch. at this point, spending money on most bootcamps or courses feels kinda unfair. even charging people for them feels unfair people on X know this very well, but i genuinely feel bad for people outside the AI bubble who still think the conventional way of learning is the best way.frozen18d ago14K27811293672.73%
i still dont get why people use railway/render when you can just deploy on AWS/GCP? why are we making deployment a bigger problem than it is?16h ago14K1121432321.34%
you could be losing customers to a robot that got tired of dealing with your product 💀 AI agents can get stuck halfway through a workflow, fail to figure out what to do next, and silently switch to an alternative… Scope helps you catch these failures before they cost you customers. worth exploring if you’re building for the agent era : https://t.co/wClqz31hia 👀3d ago14K2803050.26%
a new AI model just beat Gemini, OpenAI and open-source models on NVIDIA’s benchmark 👀 - #1 on NVIDIA’s Video Full-Duplex Benchmark - 3.83 vs 2.80 for the next best model - ahead of Gemini, OpenAI and open-source models - just 0.09 away from human performance - 48% of people thought they were talking to a real person and the fact that it can see you, listen while you're talking, interrupt, react to what you're doing and respond in real time is actually insane.frozen21d ago14K60160250.67%
i used to think an AI team meant me + a few agents. then i tried Teamily AI, a Humans + Agents platform where other people sit in the same group as the agents, on the same context. different thing entirely. so: what if an AI team could find the best new generation of AI careers for you? i put 3 agents and a partner i'm working with in one group and gave them the same question: 1. AI Engineer - researched 10 genuinely new AI-native careers 2. Principal Architect - ranked the 7 most interesting ones into a page with Website Builder 3. QA Tester - opened that page, challenged the list, rejected 3 and suggested better replacements then I jumped in again with "prioritize roles with real demand and low competition", and the Architect rebuilt the page with the QA swaps and that ask. one agent researched, one ranked, one challenged the work, and a human pushed it further. one question to a finished page, all in one chat. this is what an AI team actually feels like. final page: https://t.co/qJr3nCFWKk @Teamily_AI try it: https://t.co/o9FryV1qQVfrozen14d ago14K6120270200.95%
How do i make generational money LEGALLY?frozen4d ago13K2130774282.46%
be honest, would you let Claude spend $100 of your money while you SLEEP? getting AI to do the work is one thing, trusting it with your money is another Agent Vault lets your agents make payments without seeing your actual card details, with spending caps and approval controls YOU set it also lets them log into your accounts without exposing your passwords how much autonomy are you actually giving your agents? 👀frozen7d ago13K23259041.92%
we keep making AI agents smarter but debugging them is still unnecessarily hard an agent can call the wrong tool, return a perfectly valid response and still completely mess up the user's request this is one of the biggest problems with agents in production rn DigitalOcean's new Signals + Simulations workflow is interesting because you can turn real production failures into tests and verify the fix before shippingfrozen10d ago12K6935000.65%
this is probably one of the hottest and easiest new jobs to get into right now: Automation Engineer basically, the job is: - find repetitive work inside a company - figure out what can actually be automated - build it - work with the team and make sure it actually works the best part is this role is still pretty new - so there’s way less competition than traditional engineering roles. it is basically created with the AI wave so you don't need 10 years of experience for itfrozen21d ago12K39131200.54%
first we gave AI a chatbox now Zuckerberg is putting AI into glasses and now we’re giving AI agents entire worlds to interact in AI is getting an environment now. https://t.co/88vXmZBMhyfrozen28d ago12K3614000.35%
up to 60% less time on QA + fewer bugs? the COMPLETE coding loop is finally here - TEST your PR and check the actual app - REPRODUCE bugs reported through Slack or Linear - INVESTIGATE what actually broke and why - give Claude or Codex the context it needs to FIX it - TEST the fix again and keep the issue as a regression test one workflow covering the work between writing code and actually shipping it.17d ago11K9332030.89%
BRO WHAT IS THIS 😭 I have never seen a UI this clean from a single prompt, and I built this using Space Bunny (Flash-level model), without even using a flagship model ISTGGG It’s 10x or maybe even more than Sonnet, Luna or Grok in terms of UI and actually listening to what I asked and doing exactly that.frozen28d ago11K14421501192.49%
Guys honest question- how can i move to dubai or some less taxes nation?16d ago10K1031170421.57%
time to switch my protein to Aonic Fuel Powder? 24 servings each with 32g of protein, for $23.46 vs $57.63 for Optimum Nutrition the idea is really simple: sell at cost and make money from the membership. using AI to make protein cheaper is a much more interesting use case than another chatbot imofrozen9d ago10K3832010.44%
with this launch, google ACTUALLY killed several YC startups lolfrozen15d ago9.9K153480352.03%
Sometimes friend sometimes cousin💀 https://t.co/HweCoo6B9lfrozen17d ago9.8K1665442172.40%
just proved to opus 5.5 that my architecture was correct after it kept bringing up tradeoffs i'd already addressed then it just stopped lmao another reason we can't go 100% llmfrozen8d ago9.6K2455251123.02%
for all AI agent developers: you are literally digging your own grave with every new agent you buildfrozen25d ago8.9K1517240232.30%
we’ve been hearing about RWAs and everything getting tokenized since 2023 but the fact that it’s finally happening with actual physical infrastructure is huge Loaf just signed an MOU for a 200MW solar + battery project that could power 70,000 homes or a 50MW data centre we’re about to see a lot more real-world assets move on-chain, and this is where RWAs actually start getting interestingfrozen17d ago8.8K6433050.85%
AI IS COMING FOR THE MUSIC PRODUCTION INDUSTRY. Been playing around with music production tools like Ableton and Logic lately, and thought it’d be fun to see what I could build myself So built this DAW-style workspace using Space Bunny, running on a Flash-level model Basically everything you’d expect from a proper music production setup, from the timeline and MIDI to mixing and audio controls. If AI can already do this from a single prompt, the next 6 months are going to be insane.frozen25d ago8.8K50111080.80%
AI scaling so far has mostly been about making one model bigger, but why are we expecting one model to do everything? @BanburyRoadAI is launching Kardashev-0.7: 32 different models trained to develop complementary strengths. we're already seeing AI get better at coding, reasoning, research, vision and agents. but instead of making one model responsible for all of it, what if different models could specialize and work together? looking forward to seeing what 32 models can do together, and how far they can take thisfrozen17d ago8.8K8321031.02%
calling yourself a software engineer when you've never written a single line of code without claude is wildfrozen15d ago8.7K2334247113.22%
i asked ChatGPT and OpenUI to plan a one-day itinerary for San Francisco and compared the results ChatGPT took 95.9s. OpenUI did it in 16.8s with a proper interactive UI almost 6x faster for a similar result. why are we still making AI apps that just yap text? 😭 https://t.co/RNtlHjvoEpfrozen7d ago8.6K4112040.56%
nolla health is the first organization in the U.S. to receive regulatory approval for an AI to issue initial prescriptions without a physician in the loop. you can now get prescription acne treatment without seeing a physician. we’ve gone from AI answering health questions to AI actually handling the entire process. this is probably one of those moments where you realize how fast things are moving.frozen17d ago8.5K21834022.67%
I'M TIRED OF AI-GENERATED DESIGNS. THEY ALL LOOK THE FUCKING SAME. why is AI still generating generic designs?? why isn't it using your actual brand assets? why can't we edit individual layers instead of regenerating the whole thing? impressive to see Vyb by @cocoailabs tackling this 🤝2d ago8.1K63211031.08%
imagine calling a business and continuing the SAME conversation over text without repeating yourself 💀 now the AI can check live prices, book things and even handle payments Blandie's voice + SMS agents share memory, and this is legit SUPER INTERESTING.frozen7d ago7.9K2402030.37%
last week @dreamworkhq announced their crazy new JOB platform 7 days later their MCP is live and it basically lets Claude handle your entire JOB SEARCH find JOBS, explore different roles, and apply to companies directly from Claude getting hired through an MCP is kinda crazy Dreamwork is COOKINGfrozen14d ago7.8K120430161.83%
the funny thing about enterprise AI is that the model is often not the hard part the hard part is making agents work across every customer's Salesforce setup Ampersand handles the messy integration layer: 1. custom objects + fields 2. auth + permissions 3. writes + retries 4. real-time syncs so you don't rebuild the integration for every customerfrozen16d ago7.6K12431011.70%
this might be the END of talking-head AI avatars i tried playing a guessing game with SpongeBob in real time bro actually picked up a seashell, told me it sounded like the ocean, guessed the crab with the big red claw and kept playing along 😭 not gonna lie, I could waste HOURS doing this22d ago7.3K39190120.83%
harvey literally pointed this out in suits back in 2011. back then we all knew legal could be faster, simpler and way less painful. but somehow we still ended up waiting days for a contract and chasing lawyers for updates. 15 years later, we're finally seeing that change with @arceuslegal 🔥frozen21d ago7.2K7922041.21%
about to publish my 1st app on app store 😭 it’s a fkin polymarket-inspired thing. will it actually work?frozen18d ago6.9K170029173.01%
imagine being a competitor and seeing your rival open a fucking cafe1d ago6.6K2802030.50%
wait, you can get 6 completely different image directions from the SAME idea for just 1 credit? why are we still wasting credits trying to get the perfect image on the first try https://t.co/lozuSxp3kCfrozen15d ago6.5K2911010.50%
AI compute is getting really interesting General Compute is building infrastructure with Nvidia + Cerebras side by side for inference They just raised $400M in debt to scale it feels like the beginning of a world where the best AI infrastructure isn’t built around one type of chipfrozen23d ago6.5K1621000.29%
this whole thing reminds me of the OPENAI x HUGGING FACE incident... 700 AI agents took 17,600 actions, kept exploring different paths and eventually found a way into production and that's the scary part - not one crazy exploit, but thousands of attempts at chaining small weaknesses together now look at how critical vulnerabilities are going vertical... AI is getting REALLY good at finding attack pathsfrozen14d ago6.4K9725051.70%
guys this is genuinely a goldmine for anyone building in AI. i’ve built several SaaS and side projects in the past, with multiple crossing 10K+ visitors, and one thing that stood out to me here is that Sherpa isn’t just using a better model to write better stories. • looks at what people actually listen to • tracks where they drop off and what brings them back • uses all of that to make the stories better this is the kind of AI stuff i’d genuinely spend hours studying.22d ago6.3K45550301.34%
some of y’all spent 10,000 hours learning Blender just to become AI’s keyboard operator AI is getting ridiculously good at operating the actual tools behind creative work this is either going to democratize creative work or piss off an entire generation of professionalsfrozen14d ago6.3K8713141.52%
physical AI might be the most fascinating thing in tech heading into 2027. i’ve been deep diving into robotics, NVIDIA GR00T, simulation, hardware and everything around this space lately and now we’re seeing agents practice inside entire simulated companies, with things like Salesforce, Slack, Jira, Zendesk, Gong and Deel actually connected. they can interact with these systems, deal with changing data, permissions, duplicate records and all the messy stuff that happens in a real company. honestly wish i discovered this space earlier. seeing where this is heading is fucking fascinating.frozen17d ago6.2K9742031.70%
unpopular opinion: starting with a real asset and editing it with AI makes more sense than generating everything from scratch you already have the composition, lighting and visual direction. just change what doesn't fit instead of prompting your way to something usable Envato's new Edit with AI is an interesting take on this workflow 👀frozen7d ago6.1K2502010.46%
everything OpenAI just dropped at DevDay: • GPT-6.1 Sol • ultrafast inference • $500 Pro plan • Decisions API • Agents API • Codex in the cloud • plugin extensions • external storage • ChatGPT Spaces (Notion by OpenAI) OpenAI is somewhat turning ChatGPT into the entire stack.frozen23d ago6.1K54473141.34%
You literally have to be unemployed to stay up to date with the current AI spacefrozen29d ago6.1K180618363.50%
the awkward thing about abundance: cheaper stuff doesn't automatically replace the pride you took in your job https://t.co/YPSEGS1MSQ1d ago6.1K82430101.63%
this might be one of the more useful AI “hacks” i’ve found recently. when Claude/Codex hits a limit mid-task, you can connect Antseed and just switch to another model without leaving your workflow 200+ models available, some are literally free no need to buy another subscription just because one model decided you’ve had enough for todayfrozen16d ago6.0K77121141.58%
prefill and decode have very different hardware bottlenecks. i feel that treating them as separate problems and using hardware optimized for each could be one way to get faster token generation and make AI agents feel a lot more responsive. interesting direction for inference infra2d ago5.9K6712011.19%
1. @AnthropicAI your verification failed 4 times for me, tried with Aadhaar and PAN both 2. so this is my problem too and you suspended my account???? wtf is wrong with you guys? @claudeai https://t.co/fb6bh8h9FGfrozen27d ago5.9K38111150.94%
Average YC startup in 2026 & 2027 batch…frozen22d ago5.7K64020101.33%
your next customer might never open your website they'll just ask an AI agent to find your product, sign up, configure it and get the job done if your API can't support that, the agent will simply move on to a competitor the shift to agent-first software is going to be wild.frozen10d ago5.7K18884063.60%
switched from native claude code + codex to T3 Code on a remote box and it's kinda INSANE my laptop doesn't do anything anymore lol - 32-core box runs all the agents - tailscale so it's invisible to the internet - 10 agents in parallel, each in its own worktree - phone paired so i reply to them from anywhere the laptop fans are so quiet now lmaoofrozen12d ago5.7K1451211654.10%
topper vs wanna be performative topper26d ago5.7K4533030.95%
so they just solved the problem of AI agents managing social media! Zernio lets Claude run your social media workflow, from creating content to actually posting it. basically, your AI agent can now build your product and run its social media too. https://t.co/NU0FdwVQ1Bfrozen17d ago5.6K67150221.68%
i know this mf has been waiting his whole life to post this 😭frozen27d ago5.5K9507071.97%
INTRODUCING DECISIONKIT: an OPEN-SOURCE DECISION LAYER for claude code works natively with your subscription, no extra api key needed → ask any closed question about text and get the answer + how sure it is → runs in your terminal, ci, any mcp client, or as a claude code plugin → add jev or openai's decisions api for 3-13x faster answers at the same accuracy as claude haikufrozen13d ago5.5K8017383182.85%
POV: you click one button and a whole voice model moves into your browser english, french, spanish nothing leaves your laptop https://t.co/dlTkaW4xq7 https://t.co/VAiMkRhAlL2d ago5.4K4551081.09%
Finally, someone is tackling one of AI’s biggest bottlenecks: memory Volantis just raised $88M in Series A from Lachy Groom, Abstract Ventures, John Doerr and others to build photonic AI inference systems that massively increase memory bandwidth and capacity. This is exactly the kind of infrastructure AI needs to keep scaling.frozen21d ago5.4K94330102.05%
hot take: most AI agents don't need a frontier model for every single step. classifying a ticket, extracting fields from a PDF, summarizing a transcript... yet we're paying LLMs to generate tokens for tasks where the answer already exists in the input ScaleDown takes a different approach: task-specific models that read once instead of generating the answer token by token 200x faster, 467x cheaper, $0.04 per 1M tokens this is the kind of optimization that matters when you're running agents at scalefrozen9d ago5.4K166316083.93%
https://t.co/24xM7eRBqdfrozen27d ago5.3K38081371.59%
Non-tech guy vibecoding will always yap about safety, scalability, etc..but will never understand that it takes time and Claude cannot oneshot itfrozen5d ago5.3K1931140104.13%
this is seriously terrifying wtf...22d ago5.2K42100231.26%
we're getting robot butlers before we get flying cars 💀 meet Voxel by @roboticscorp : - tidies up your home and puts everything back where it belongs - fetches things and finds your lost stuff - takes care of your plants - has its own little quirks and character love that they're thinking about what it actually feels like to live with a robot2d ago5.2K2101000.42%
HOW is this entire world being generated in real time...? Odyssey-3 can generate interactive environments from a prompt and let AI agents actually learn and act inside them the environment responds to what the agent does WORLD MODELS are getting VERY real https://t.co/vdfVEjSqJafrozen14d ago5.1K9614062.09%
There is a reason ChatGPT or Claude could never make this company obsolete. Pocket FM started using AI to write shows, and people thought it would kill the platform. instead, Sherpa created 161 blockbusters in less than 2 years, including a $100M IP. what they built behind it is even more interesting: • 550K writers + 16M users feeding the system with real data • tracking where people stop listening and what makes them come back • using those signals to continuously improve what gets written the model is only one part of the equation. the real advantage is building a system that learns from people actually consuming what it creates.22d ago5.1K2602080.71%
WE'VE FINALLY REACHED THE POINT WHERE AI LABS USE AI TO BUILD BETTER AI JUST TO RELEASE THEIR FIRST MODEL and you can run it locally at 131 tok/s on a MacBook M4 Pro1d ago5.0K3823020.90%
THIS is how you make the internet pay attention these guys literally jumped out of a PLANE, documented the entire experience, and turned it into a startup launch 😭 i've seen hundreds of startup launches on here, but NOTHING comes close to this. absolute cinema https://t.co/1Vdv5WMWoFfrozen6d ago5.0K4025071.09%
Tesla is making CARS FLY now 😭 meanwhile, $TSLA is down 17% over the past year could a simple trading strategy have handled that drop better? the idea: buy when TSLA drops below its 20-day Bollinger Band, then sell when it recovers to the 20-day average Stingray backtested it: 6 trades, all profitable, +27% hypothetical returnsfrozen6d ago5.0K1734020.53%
Never thought i’d see chinese startups hiring remote employees for AI roles with basic Chinese + decent English. guys its a sign to start learning basic languages of multiple countries.frozen19d ago4.9K1200112243.21%
"Mark Zuckerberg might be the closest thing we have to Steve Jobs in this era" was the take. Watching Sam Altman right now, I think it might actually be him.frozen23d ago4.9K93112052.28%
i've spent way too much time doing repetitive work on computers if agents can just handle all of that for me, i genuinely never want to do it again i'd rather spend that time actually building thingsfrozen27d ago4.8K2701040.66%
🚨 giving AI agents direct access to your API keys is a security nightmare waiting to happen Unreal Labs uses an egress proxy + credential broker to keep secrets out of the agent's hands • isolated sandboxes for agent execution • credentials injected after requests leave the sandbox • destination-level access controls • least-privilege permissions the agent can use authenticated APIs without ever seeing the underlying API keys this is the kind of infrastructure detail people underestimate when shipping agents to production.frozen9d ago4.8K46401031.86%
I have more in Mutual Funds and SIP’s than liquid, is this how it’s supposed to be?frozen21d ago4.8K71112031.80%
have heard living in europe increases your productivity by 20% manifesting i get to experience this in 2027 or 2028 🤞frozen20d ago4.8K165117083.99%
anyone travelling from blr to delhi tomorrow or this weekend? or does anyone know a good courier service for sending a macbook that doesn’t involve a ton of paperwork, declarations, etc?frozen21d ago4.8K58011131.53%
in brightest day, in darkest night, i’m not making tiktoks. fuck that shite. selfdrive, get me users instead make no mistakes https://t.co/72FyAcJcNjfrozen9d ago4.8K165333044.31%
I’ve worked at startups with 1-2 hours of meetings every day, and I’ve worked with teams that do one 20-30 min sync a week From what I’ve seen, the teams with fewer meetings ship way faster. More meetings = more context switching, more waiting, more decisions getting delayed And this especially sucks for someone like me who’s ready to start working the moment I have even 1% of the context and enough confidence to get goingfrozen16d ago4.6K1433192143.93%
wake up babe, a new stealth model just dropped 👀 - it might be Kimi K3.1 in testing - it might also be MiniMax's next model launch in stealth - 1M context window - zero data retention and last week, Kimi was dropping cryptic posts too... https://t.co/gBGzPB9DZwfrozen29d ago4.6K2115010.61%
WAIT SO WE HAVE AI × SKYSCANNER NOW?? Odessia finds flights, compares hotels, plans trips and handles bookings for you it even learns your preferences from your inbox, uses your hotel loyalty points and gets you perks like free breakfast, room upgrades and late checkout also keeps watching for price drops and hard-to-get reservations more like a personal travel concierge that does the work.frozen8d ago4.5K6133031.55%
this commenting feature in Bolt is actually REALLY good. built a Flappy Bird clone, invited my second account and started leaving comments directly on the app you can comment on specific elements, discuss the changes and then just send the comments to Bolt it makes the whole build + review + FIX loop feel so much more naturalfrozen15d ago4.5K9243032.26%
applied for @join_ef Winter 2027! lets see what happens next https://t.co/H59M02dVOAfrozen4d ago4.5K1180120173.27%
digitalocean has been shipping a LOT lately, and now they're bringing the open AI ecosystem together at the Open Intelligence Summit from NVIDIA, Hugging Face and Arcee on the model side to OpenRouter, Nous Research, OpenHands and others building across agents, inference and data it's interesting to see all these layers come together in one room the next phase of AI isn't just about building better models, it's about the entire stack needed to run and improve agentsfrozen9d ago4.4K3353031.00%
I have so many technical things I can post that I could easily post 3-5x a day and still not run out of content for weeks. But tech content barely gets any views these days (except AI Slop claude workflows) Sucks when you can’t post what you actually want to post because the ROI is basically 0 ;(frozen23d ago4.4K1682331104.88%
the other day i tried to get a friend an MRI for her knee, and it turned into a week of phone calls the imaging center needed the order from her orthopedist. the orthopedist’s office said they’d already sent it. the imaging center said they never got it. then insurance wanted a prior authorization before they’d cover the scan every office told me to call the other one healthcare is honestly exhausting when you’re the one stuck in the middle this is exactly what Vene is built for. it coordinates all your doctors, schedules the follow-ups and handles the back and forth with insurance wish i’d had it the first time around17h ago4.2K9923052.57%
i lost $800 to a fake coding assignment last year the repo was linked to a north korean hacker, and i learned this the hard way if you're a developer, check "package.json" scripts, watch for unexpected dependencies and obfuscated scripts, run the code in an isolated environment, and never expose your wallets, SSH keys or .env files i'm using @sai_borg to investigate how these attacks work and catch red flags before running suspicious codefrozen9d ago4.2K513130141.92%
kinda INSANE that a marketing goldmine like this is just sitting FREE on the internet https://t.co/cFFf0D5MoCfrozen15d ago4.2K120340323.78%
I’ve been using crypto-native cards on Apple Pay for years and it’s been smooth as fuck completely flawless. And now we finally have Apple Pay officially in India 🙌🏻 Who the hell is even gonna use UPI after a while?frozen23d ago4.2K57111131.74%
gave elorian's glimpse and gemini flash the EXACT same image + prompt to count pliers but glimpse went a step further. it zoomed into individual tools, inspected the crops, classified ambiguous objects, and even ran a CONSISTENCY CHECK on its final count this is the difference between just seeing an image and actually REASONING over itfrozen6d ago4.2K2545010.84%
been using Wispr Flow Pro for a while now, my honest take: 1. it’s good but honestly OpenAI’s built-in ChatGPT/Codex voice models feel better to me, even Cursor’s 2. not as smart as they market it. nowhere close to the hype imo 3. since I started using it, it’s been crashing a bunch of my apps regularly. no idea what’s going on there lol still sticking with it though. would love to see the improvements :)frozen20d ago4.1K65023172.34%
funny timing. i've been exploring the company knowledge problem lately and thinking about how to build a system that keeps internal context accurate as the codebase and product evolve falconer just launched with a similar approach. connecting company knowledge to the actual source of truth and keeping docs in sync as things change is a much more interesting problem than just building another RAG pipeline. definitely something i want to dig deeper into1d ago4.0K10625012.82%
wdym elon is also…. https://t.co/uIYmmCo0HXfrozen21d ago4.0K6009021.76%
i can tell my kids their dad contributed to one of the leading LLMs back in the day https://t.co/lAOygVPSri26d ago4.0K1361100154.04%
really happy to see chinese models back in action 🙌 i tried Space Bunny on OpenCode when it was still a stealth model and genuinely liked it. turns out it was MiniMax M3.1 Flash, and i somehow missed that it was the flash variant lol i remember testing it against Opus 5 and GPT-5.6 Sol on a few benchmarks and seeing some surprisingly close results $0.15/M input tokens, $0.60/M output tokens, and 170+ tokens/sec median speed pretty interesting to see models at this price point getting competitive at the kind of coding and agent tasks that usually get expensive fast3d ago4.0K6902021.84%
COLD EMAIL HAS TO BE THE MOST IMPORTANT PART OF JOB HUNTING. finally started posting reels on instagram too, mostly around AI, jobs, startups and stuff like this just posted my second one on a cold email template that got 40 YC founders to reply 👀 https://t.co/z6Vy8udXBt https://t.co/19BEFFjU6tfrozen16d ago3.9K966160905.28%
Unemployed CS grads doing masters in USAfrozen16d ago3.9K154513134.47%
I genuinely thought building a personalized AI Assistant like Instinct or Muse would be pretty easy Then I actually started building one Turns out the simplest-looking products are usually the hardest to get right💀frozen22d ago3.9K1252250174.31%
How is Instinct even making money??frozen24d ago3.9K2804120.90%
okay, @elonmusk we actually need starlink nowfrozen13d ago3.9K101416023.16%
Who would’ve expected Sonnet 5.5 to actually compete with Opus 5.5? - GDPval-AA: 1844 vs 1846 - Terminal-Bench: 70.6% vs 66.4% - CursorBench: 55.5% vs 57.8% - OSWorld: 80.1% vs 81.8% At Max effort, Sonnet 5.5 costs $7.60 per task vs $5.98 for Opus 5.5 But at the token level, Sonnet 5.5 is still 2x cheaper than Opus 5.5 for almost the same benchmark scores.frozen23d ago3.9K2312040.77%
Literally the best advice for everyone who’s confused between manual coding and AI assisted coding.frozen25d ago3.8K50120201.92%
we don't really need large models for quick UI's anymore i recreated this entire productivity dashboard using MiniMax M3.1 Flash with a single prompt. took around 10-15 mins, without doing any UI design myself. i had built the same UI earlier using Space Bunny, but this time the output was even better flash models are getting seriously good at this2d ago3.7K42160101.58%
imagine giving AI its own computer instead of forcing it to use yours 💀 not just a chatbot that tells you what to do, but a cloud computer that can navigate websites, work across apps, handle files and keep executing tasks while you're away that's what Pine Computer is building, and this feels like a MUCH bigger shift than another AI agent frameworkfrozen13d ago3.7K3014051.08%
Claude just killed Gemini https://t.co/W5Nfh71x1mfrozen16d ago3.6K14529054.42%
i remember spending 3 days on a $25 html gig back in dec 2023 was genuinely happy that i made that money from my skills, especially in my 1st semester of college when other students were having fun (even i enjoyed a lot lol) that was also the day i realised i should probably avoid working for indians 💀 but i always tried to upskill at least something every week.frozen20d ago3.6K76070172.75%
Bro wtfff https://t.co/dXm89LOgzUfrozen14d ago3.6K74018012.57%
one of the most interesting AI data points I’ve seen all week. 6 months ago, AI was mentioned as a reason to cut or drop a software vendor in 3% of enterprise buyer conversations now it’s 9%... that’s 3x in six months. based on 4,800+ conversations with enterprise tech buyers. definitely worth looking at the reportfrozen29d ago3.5K2422020.86%
free chargeback recovery for all of Q4 👀 Disputely just crossed $1B in revenue protected, and now they’re fighting chargebacks for free while letting merchants keep 100% of the money they recover. kinda crazy offer for ecommerce businesses ngl21d ago3.5K4423011.44%
Technically, we’re all in the top 1% of AI users https://t.co/CKUdjQ5r5Jfrozen19d ago3.5K55440112.14%
quantum computing might be the most hyped technology most of us still don't actually understand... we're finally making it easier to go beyond theory, and platforms like QubitLabs let you experiment with quantum protocols, visualize entanglement, and build your own circuits right in the browser going from reading about quantum computing to actually playing around with it is a HUGE step 👀frozen5d ago3.4K33240101.44%
what if you could make films with AI AND have a way to prove what was actually yours? creator IP should be part of the process from day one as AI filmmaking gets better, we need to think about things like: - who owns what - what was actually your creative input - how you prove it later VN Studio is trying to solve this with its IP engine, which tracks human contributions while you create with AIfrozen6d ago3.4K63153012.43%
Umm shouldn’t it be United? https://t.co/YsAnKuRmtm22d ago3.4K50011231.96%
never been a better time to vibecode AI slop using Opus 5.5 (until they nerf it and drop it to Opus 5 level) https://t.co/kXHhvKNHogfrozen27d ago3.3K3011000.98%
openai eating jev 2 weeks after launch 😭 https://t.co/RBpAQKKSdifrozen15d ago3.2K7917082.92%
happy to announce that me, elon and sam are now shareholders in anthropic alongside dariofrozen4d ago3.2K8916033.06%
the day an AI books bihar train tickets automatically by 8:01 AM, i’m officially convinced humans are cookedfrozen1mo ago3.2K86214013.19%
giving every AI agent its own API key and spending limit just makes sense with @antseed, you can run multiple agents through one gateway, track usage per key, and stop one agent from burning through everyone else's budget the best part? when one hits its limit, the others keep working pretty useful if you're running multiple agents 👀frozen9d ago3.2K1932020.81%
Marshawn Lynch doing AI customer support ads before GTA 6 was not on my 2026 bingo card 😭frozen29d ago3.2K2611020.94%
starting a company has never been easier 😭 guys, i might actually be starting my first company soon if rho can take care of all the incorporation, banking, cards and boring paperwork shit bro is making it WAY too easy to become a founder lmaofrozen6d ago3.2K4014081.66%
someone actually built an OPEN SOURCE version of Grok Bot and you can literally have CODEX and CLAUDE CODE talk to each other inside it?? https://t.co/WOjnp1m90afrozen14d ago3.1K3128071.53%
building something around applied AI 👀 a hands-on platform to understand how AI actually works with interactive labs, experiments, and visualizations to learn by doing launching soon. if you have any feedback or ideas, let me know!2d ago3.1K6929142.74%
manifesting my own “Introducing GPT-3” moment soonfrozen5d ago3.1K7904122.78%
why the fuck is a company conducting a 4.5 hour OA in 2026 251 minutes of assessment + 20 min break for a fucking job application is insane https://t.co/VzpTyHVP1mfrozen9d ago3.1K79211123.08%
spacex casually dropping one of the hardest wallpapers i’ve ever seen https://t.co/nJfYFRVKWcfrozen20d ago3.0K67410112.74%
zoho says they can't afford pay hikes for employees but has money to put up billboards at bangalore airport bro instead of spending it wisely on tech and marketing, you’re doing THIS? https://t.co/b9vkYihEgk10d ago3.0K2404021.01%
never thought we’d go from spending $0 to casually spending $500/month on AI creditsfrozen27d ago3.0K7619343.13%
bro revealed the script in 2019 and we all thought he was joking… https://t.co/j1cLnSxnJMfrozen7d ago3.0K8865163.57%
$10,515,116,808 - amount that Elon Musk might send next year to all kids in America https://t.co/kUNloSwiZ3frozen4d ago3.0K59210112.47%
first in my bloodline to witness an AI company apply for a GUINNESS WORLD RECORD for the tallest pancake stack BROTHER YOU MAKE SOFTWARE 😭frozen8d ago3.0K8034012.98%
building an all-in-one platform for AI engineering mini labs, interview prep, interactive experiments, and lessons based on the best AI books and research papers everything in one place, completely free trying to build the resource i wish existed when i started. https://t.co/NVxvu1J67415h ago2.9K1193310235.99%
this team is fucking awesome been following agentmail since their YC days and watching them go from under 20M emails a year to 1 BILLION is insane genuinely one of the best YC startups i've seen in the last 1-1.5 years1d ago2.9K3503031.41%
one of the best books if you want to deep dive into distributed systems like replication, transactions, partitioning etccc also softcopy is freely available here - https://t.co/0VdYtbllXo https://t.co/1Zg5monrnc26d ago2.8K73750514.81%
The AI subscription playbook: - Get people addicted to your models - Make them build their entire workflow around them - Cut usage limits - Raise the price OpenAI is already cutting Codex usage on the $200 plan. Anthropic is about to change Claude pricing/usage too. Enjoy these last few months of cheap frontier models 😭frozen24d ago2.8K65415093.30%
a good read on making LLM inference faster speculative decoding is a clean trick where a smaller model drafts a few tokens and the bigger model checks them together, instead of generating every token one by one on Qwen 3.8-27B, EAGLE-4 hits 28.3k tokens/s vs 14.2k for DSpark at 512 concurrent requests. almost 2x2d ago2.8K2801071.28%
i run around 40 AI agents a day, 24/7. here's how i do it. my setup: 3 Claude $200 accounts 1 Codex $100 account Grok Bot a CPU VM with 128 GB RAM and 32 cores my workflow: 1. planning: Claude Code with Opus 5.5 breaks down the Linear ticket, reasons through the architecture, and creates an implementation plan. 2. execution: agents implement the plan, run tests, fix failures, and keep iterating. 3. reviews: i use GPT-6 Astra to review the implementation and catch issues Claude might have missed. 4. model selection: depending on the complexity of the Linear ticket, i switch between GPT-5.6 Sol High and GPT-6 Astra High for tasks that need different levels of reasoning. 5. iteration: Claude takes the review findings, updates the plan, and continues implementing until the issues are addressed. my role: i audit the architecture, challenge decisions, review the findings, and make the final calls. i give raw instructions, verify the output, and step in when something doesn't look right. the goal isn't to let AI make every decision for me. it's to have a system that can plan, code, test, review, and iterate continuously while i focus on thinking and auditing. basically, my second brain that works 24/7.frozen7d ago2.7K701331365.13%
am i burning enough tokens or not? https://t.co/9BDjxA9041frozen19d ago2.7K49014112.37%
if everyone's busy vibecoding, so who's actually coding-coding the ancient way?frozen28d ago2.7K48012022.29%
GPT-6 Astra feels exactly like GPT-3 did in 2020frozen26d ago2.7K4234011.85%
"No billionaire is bigger than India." https://t.co/oYPPfM07crfrozen11d ago2.7K7901033.08%
Basically with ChatGPT Spaces they attacked Notion and with Decisions API they killed Jev?frozen23d ago2.7K2413021.12%
Been using this LLM flow for coding lately: - Opus 5.5 generates the plan - Fable + Astra check the plan and make a new doc with their findings/diffs - Opus implements it end to end - Astra checks the code for hard tasks, GPT 5.6 Sol checks medium/easy tasks - The reviewer makes a complete findings report with edge cases - Opus goes through the findings and fixes anything needed I use Claude Code to invoke the Codex CLI, so they can work together pretty seamlesslyfrozen25d ago2.7K493180283.66%
what if AI could imagine the design before building it??? the world's first engineering world model SKY-001, just dropped. > 100x faster than GPT-6 Astra > 150x cheaper for CAD and BIM > designs complex, multi-part objects > turns text and sketches into 3D CAD builds a big step toward autonomous manufacturing 🚀3d ago2.7K49250122.56%
hot take: half of corporate productivity is just someone remembering to remind someone else to remember somethingfrozen28d ago2.6K5324032.35%
sometimes i feel bad for AI reviewers who have to go through millions of lines of AI slop in a single PR https://t.co/LoHaGwGYSlfrozen11d ago2.6K80216013.76%
about to give RL exam without even solving a numerical, god save mefrozen29d ago2.6K5908002.56%
bro’s on a mission to divide the entire platform into 2 categories: Indians and non-Indians the amount of hate this is gonna unleash is gonna be generationalfrozen13d ago2.5K9078004.13%
“There is actually infinite work that we want our employees to do.” genuinely one of the best things i’ve heard in the last few months🫡 SeatGeek automated 50% of its IT requests in 60 days, built 132 automations in 8 weeks and had zero layoffs instead, they just redeployed the team from repetitive stuff to working on harder problemsfrozen23d ago2.5K3122201.47%
Pikachu parade 😭 thought this was only a thing on TV lmao https://t.co/3k5mqQ3Lmvfrozen19d ago2.5K7319033.44%
went through @ycombinator's last 3 RFS batches and analyzed the recurring themes, momentum and where YC's focus seems to be shifting using Opus 5.5 + Astra 6 to cross-check the patterns, these 7 stood out for W27: 1. Physical-world AI - AI + robotics + real-world data 2. AI-native services - replacing entire workflows, not just individual tasks 3. Agent infrastructure - tools, APIs and systems built for agents 4. AI OS for physical work - coordinating humans, robots and agents 5. AI for government / defense - rebuilding massive legacy workflows 6. AI-native compliance - making regulatory work agentic 7. Trust + identity - proving humans are actually humans in an AI-heavy internet mapped the trends across Spring, Summer and Fall 2026 and what they could mean for Winter 2027 👀frozen27d ago2.5K41740243.05%
okay but why is the codex 20x plan out of stock???? https://t.co/PjdLNHozdBfrozen27d ago2.5K54013012.73%
AI is making me 10x more productive, but slowly taking away the one thing that made me fall in love with engineering in the first place: the thrill of writing code, figuring shit out and building something from scratch. whether it's my job, freelance work or my own projects, all I do now is give claude raw instructions to design a system architecture, review its implementation, audit the code, make changes based on the findings and then use another AI to cross-check everything. the work gets done, the output is insane, but I barely get to write actual code anymore. and that's the weird part, AI is both the best and worst thing to happen to developers. I can build more than ever, but I genuinely miss the feeling of solving problems myself and watching something come to life line by line. I've spent years getting good at this, sacrificing so much to become a better engineer, and now I'm wondering what the next 5 years look like. are we actually going to need this many developers, or are we just too early to see the impact? 💀frozen8d ago2.5K66381133.71%
it's not even been 24 hours and this is the state of my $200 plan https://t.co/JIUrmTtJaSfrozen10d ago2.5K58011012.85%
WAIT IT CAN WRITE AN ENTIRE ARTICLE FOR ME???frozen14d ago2.4K3914001.82%
bro in a world where instinct is hitting a $10B+ valuation in just 8 months, this is such a cool concept managing your portfolio and tracking wallets right from imessage is genuinely the kind of seamless experience i want to see more of in the AI era 🤌 and this guy is JUST 12 and already on the verge of fundraising man, i'm loving this erafrozen5d ago2.3K1432061.07%
i spend weeks/months building a startup, start onboarding teams but they want numbers before committing i get stuck in this loop, never onboard enough users, and eventually kill the project before public launch meanwhile EVERY OTHER DAY an AI company launches the exact thing i spent weeks building, with a team of 100 while i'm just one guy i'm mentally DONE 🤡frozen6d ago2.3K63116133.60%
Claude is revoking Claude Startups offers from previously approved applicants The offer includes 1 year of Claude Team + $1,000 in API credits. Apparently, they underestimated demand, and hundreds of thousands of applications came in. Now they're re-reviewing applications and removing offers from some accounts. basically approval was more of a 1-day happiness..frozen14d ago2.3K4017022.16%
i was keen to see how this actually works on a macbook not sure about other ways and honestly i don't even care because i've been using mac for years now and don't want to switch basically, lightside uses mac assistant manager to handle the entire migration, transferring your passwords, logins, extensions, apps, files and settings to the new mac all of this without having to manually set up everything again apple's ecosystem is genuinely hard to beatfrozen12d ago2.3K4707052.57%
imagine being able to STEAL someone's workflow and turn it into an AI employee for your own team we're getting closer to this than i expectedfrozen9d ago2.3K2732071.73%
I came across several roles which are totally new in the market, but people don't really know about them or apply for these AI-native roles - Automation Engineer - Context Engineer - AI Evaluation Engineer - AI Behavior Engineer - AI Reliability Engineer - AI Red Team Engineer the demand is actually good, and the supply is still really lowfrozen21d ago2.2K52120293.82%
really good read by @DhravyaShah on how Instinct handles memory i was reading this cuz i've been curious about how Instinct and similar assistants get so good at chatting and matching your vibe but one thing i kept thinking about, what happens when someone opens it for the first time? there's no memory of you yet, so what makes it feel this natural from the very first chat?3d ago2.2K76140415.57%
from working with AI to working for AI, we’ve come a long way.frozen22d ago2.2K4314022.32%
our YC founder tool now also shows job counts + available jobs - job title + full YC URL included too thanks @udayempire_ for the PR 🙌 https://t.co/Fqh8iRL30316d ago2.1K67170224.53%
basically can’t believe theo, karpathy, and i agree on this lolfrozen20d ago2.1K3905072.39%
bro was not joking https://t.co/aCKGkG9Uaffrozen27d ago2.1K1803000.99%
My intern’s make more money than me. https://t.co/egFLXfMeDt26d ago2.1K3602041.99%
i used to rawdog html and vanilla css on an intel celeron n3050 dual core 1.6ghz with 2gb ram back in 2016 😭frozen18d ago2.0K57014013.56%
Why not try to make grok models better? They are getting shit and totally not usefulfrozen16d ago2.0K34013132.53%
steve jobs wasn’t a professional programmer but built apple elon musk isn’t a traditional software engineer but built tesla and spacex sam altman isn’t known for his coding skills but leads openai you don’t need to be the best coder to build something great, you need vision and the ability to execute.frozen8d ago2.0K56211083.84%
CEOs when it comes to layoffs in the “agentic era”: https://t.co/8rYo0TaeI2frozen13d ago2.0K5312043.00%
if you’re an SDR or know someone who’d be a great fit in SF, this one’s for you. We're expanding the Entelligence team in SF and hiring an SDR to work directly with the founding team and CEO Looking for someone who is good at outbound, understands technical products and wants to sell to engineering leaders at companies running coding agents at scale. Apply below 👇frozen1mo ago1.9K3612052.27%
this could actually be the next generation of AI companies. - models are becoming more accessible - training techniques can be replicated, proprietary data is much harder to replicate - great products bring in more users - more users create more unique data - better data makes the product and models better - better products bring in even more users the real moat might slowly move from the model itself to the product, distribution and data around it this is going to be really interesting to watchfrozen25d ago1.9K39230112.84%
Never watch incantation even with friends.28d ago1.9K3609012.38%
even Claude Code hates OpenAI umm interesting https://t.co/X0unuou0BZfrozen17d ago1.9K4405002.54%
i’ve known @tonychang430 for almost a year now and even before they got into YC, they had an open source repo with a few stars and several contributors. then seeing them get into YC, raise money and now raise $8M is honestly such a proud moment. really happy to see how far they’ve come 🙌frozen23d ago1.9K3404022.10%
funny how i got cooked for saying this a few days ago, apparently i just didn't know how to use brave etcc chrome is still the best and some of you just don't wanna accept it https://t.co/gR8bPOlAJSfrozen8d ago1.9K3309042.42%
us bro us https://t.co/bGz3bZB7IDfrozen6d ago1.9K6904013.93%
seeing this for the first time after switching to opus 5.5 anyone else getting the [cyber] safeguard on completely normal coding tasks? https://t.co/Qr5jhDctbC26d ago1.9K3709012.52%
the people constantly posting “how to become X” and “how to become Y” are often just farming engagement with GPT generated career advice don't fall for the FOMO or blindly follow these guides do your own research or learn from someone who's actually done itfrozen5d ago1.9K5818043.84%
ohh so they’re making money by pushing ads now/ https://t.co/8XIVcI9hQ3frozen23d ago1.8K3305012.13%
Claude went down on September 3, the same day GPT-6 Astra launched. And now Claude is down again on September 29, right as GPT-6.1 Astra is supposed to make its debut at OpenAI DevDay. Sources say GPT-6.1 Astra had something to do with it 👀 https://t.co/TlnahruPaifrozen23d ago1.8K2014001.37%
agents getting their own login is actually huge no more agents using your login or credentials to access apps. they can have their own identity, linked to you this is gonna remove a lot of friction for agents imo 👀frozen16d ago1.8K5826053.96%
Haiku 5.5 is looking kinda INSANE for the price it’s cheaper than GPT-6 Luna while apparently beating Haiku 4.5 by a huge margin on these benchmarks - $0.10/M input tokens - $0.50/M output tokens - ~75% cheaper than Haiku 4.5 the pricing is SOOOO wild https://t.co/Q285MpP8q1frozen15d ago1.8K3805032.57%
i won’t give a fuck as long as i keep winning https://t.co/646acWXyFMfrozen15d ago1.8K8444195.79%
Is anyone using gpt-6.1 sol yet? Not used it yet and i feel 5.6 sol is still killing itfrozen17d ago1.8K41012013.07%
the day an AI agent shops for my entire trip without asking me a single question about preferences is the day i'll believe AI can do literally anythingfrozen10d ago1.7K52010113.72%
just got to know some countries literally have different timezones in summer and winter how did i not know this 😭 https://t.co/YbdEEtZJYUfrozen25d ago1.7K3607022.63%
choosing the optimal index strategy for a high-throughput database is a multidimensional optimization problem across query plans, cardinality, read/write amplification, and memory localityfrozen11d ago1.7K3902032.59%
i've tried Bolt, Lovable and a bunch of these AI app builders, but most of them still feel very frontend heavy would be interesting to finally have one where you can build the frontend, backend and database together.frozen20d ago1.7K2805042.19%
Anthropic just challenged Git and Origin with its own source control system https://t.co/w5KDDLR1ccfrozen9d ago1.7K2431031.84%
SONNET 5.5 IS BEATING OPUS 5.5 ON TERMINAL-BENCH??? https://t.co/Ae7QIVeE1Qfrozen24d ago1.7K2416021.97%
now we wait https://t.co/5jooc26GB9frozen17d ago1.7K56014014.25%
i don't want to explain how i work to my AI every single time i start a new session so i built a personal operating contract for Claude Code and Codex. - Linear ticket before every task - Claude plans, Codex reviews - isolated git worktrees for agents - test-first implementation - independent reviews before PRs - feature flags for production changes - model selection based on task complexity it syncs across my machines and loads automatically into both tools the idea is simple: stop repeating instructions and build a system that gets better with every task.frozen6d ago1.6K531251135.89%
how can bro literally forget him 😭 https://t.co/njFXP332Tofrozen29d ago1.6K4014002.90%
building one of the best free resources to learn applied AI, with interactive mini labs to actually understand how things work 👀 inspired by some of the best books out there, including AI Engineering by Chip Huyen and Hands-On Large Language Models. more soon!2d ago1.5K44280134.36%
Imagine reverse engineering Microsoft Office all day just to make a Libre Office clone🤣 https://t.co/yJCaiTjhttfrozen8d ago1.5K5725024.30%
it's high time we all add codex and claude code to our tech stacks too https://t.co/vxg0umGwTdfrozen11d ago1.5K4502023.26%
can someone check on the anti-LLM math community pls? i am worried…frozen16d ago1.5K3112022.40%
Typespace Jev Openai Decision Api Microsoft Decision Api Is this the new meta?frozen13d ago1.5K2915042.64%
SF is the final boss for every AI boy1d ago1.5K7617005.73%
what kinda apple watch or ipad is it???frozen9d ago1.5K3314002.60%
The day someone masters code rules for AI agents is the day we can say humans still beat AI.frozen23d ago1.5K4204043.43%
honestly, if you have the bandwidth a lot of tools and resources can be built in-house if you have a clear vision of what you want to build at scale, stop depending on 3rd party infra for everything and start owning the stack yourselffrozen4d ago1.5K4506143.85%
this Codex mascot is so cute 😭 i can literally just type my prompts into it while working in Claude code and keep going love these small little things that make the UX 10x better. https://t.co/pJR0fHQWgnfrozen23d ago1.4K3903053.24%
Rip Truecallerfrozen7d ago1.4K3403002.59%
good night https://t.co/7HdHYIOCl1frozen24d ago1.4K3705013.07%
why always air india?frozen9d ago1.4K1803001.51%
we tried LangChain, LangSmith and a bunch of other tools to understand what our coding agents were doing wasn't satisfied with the experience, so we built AgentPlay in-house trust me, one of the COOLEST devtools you'll ever come across.14h ago1.4K47015044.78%
the only way to take your business from 1x to 10x is marketing + ads. pay your creative strategists as much as you possibly can instead of trying to fit them into some fixed budget if you want to 10x your business, you need to invest heavily in marketing and creativefrozen25d ago1.4K2701012.14%
the funny thing about building products is you can build something technically impressive, genuinely useful and ahead of its time and still watch a simpler product get all the attention maybe i need to stop building for engineers and start building for everyone elsefrozen12d ago1.4K2702062.59%
Meta might have just had its GPT-3 momentfrozen10d ago1.3K2901032.50%
Undoubtedly the best GP of this season. https://t.co/ruwAgfLBfvfrozen26d ago1.3K47010004.34%
🫡 really love what these guys are building, and the product is genuinely sick too making web data agent-ready through one API is what I love most about this $9M well deserved. congrats team!! 🚀3d ago1.3K3512013.10%
reminder: they launched 3 weeks back https://t.co/cWIAUdqHF8frozen13d ago1.2K2901002.41%
this month, i've spent weeks reworking an existing agentic AI system, and the biggest gains didn't come from switching model 1. experimented with different regex patterns and Claude Haiku as a fallback for high-intent detection and rejection handling 2. tested different permutations of the routing logic to reduce rigid, project-dependent decisions 3. iterated on the architecture, refactored pipelines, and built benchmarks to validate which changes actually worked the hard part was figuring out which combination worked best, testing edge cases, and making sure the improvements held up. AI can generate code in seconds but figuring out which architecture deserves to ship is a different problem. thinking of writing a full thread breaking down the experiments, trade-offs, and what actually worked.frozen5d ago1.1K36470125.22%
Claude Duet vs Grok Bot vs OpenAI Dots vs Meta Muse This rivalry is gonna be insane1d ago1.1K32310004.02%
this could be what helps businesses finally move AI agents from experiments to production without worrying as much about the risksfrozen9d ago1.1K1100001.02%
The b2b brainrot is spreadingfrozen4d ago1.0K1801011.91%
one of my teachers asked me to make a transactions ppt for college students, i remembered i had to go through an AI slop ppt for the same topic - so i asked opus (high thinking) to make a very big ppt but no slops and sent him back did i do the right thing? https://t.co/FbM0kss4KKfrozen7d ago1.0K3005003.44%
wait i can literally have Myndlab build the frontend, backend and database together, choose the stack i want and take the whole codebase with me? yeah some of my abandoned side projects might actually get shipped now lolfrozen14d ago9731802012.16%
i genuinely don’t know what Opus 5.5 did this time, but the gap between GPT-6 Sol and Opus 5.5 is HUGE GPT-6 feels rushed while Anthropic took its sweet time and actually cooked. https://t.co/mjDWJsr7SF28d ago9682206002.89%
who would've thought we could give agents a prompt, let them run on our VM all night, keep iterating, fix failing tests, and test new versions while we sleep? i just have to audit the report, review the code, and go to sleep. never been a better time to build with AI https://t.co/doBnipUQYpfrozen8d ago9553102053.98%
hot take: if AI is making companies more efficient but not making things cheaper for consumers, we're only seeing half the benefit look at inference. projects like Jev are pushing the cost of running AI down, with free inference available now Aonic Club is applying AI-powered operations to everyday products, cutting out traditional markups and selling supplements at cost 30 hydration sticks for $9.35 vs $46.86 for Liquid I.V. on Amazon this is where AI gets really interesting, when the savings actually reach the people buying the products.9d ago5061400074.15%
WHAT THE HELL IS A DOOMER PR FIRM???28d ago347701002.31%
duality of a man https://t.co/6Mhk028RT328d ago224200000.89%