← Back to leaderboard
deedydas

Deedy

Verifiedactive
@deedydas

Partner at @MenloVentures. Investor: Anthropic, OpenRouter, Modal, Wispr, Pangram, Inception, Goodfire, PrimeIntellect prior: Glean, Google Search. Cornell CS.

RosterAI Model News & ReleasesStartup & SaaS Builders
256,372 followers · 6,341 following · Last polled 1h ago
Perf. Score
83.1
Rank #13
Value Score
33.2
#184 of 268 priced
Median views
61K
p25 50K · trailing 7d
Followers
256K
+0.2% 7d
ER (impr.)
1.83%
eng ÷ impressions
ER (foll.)
0.79%
eng ÷ followers
Est. CPM
$57.01
post-rate basis
WoW views
-47.2%
▼ falling

Rates & value · est. from 61K median organic views

overpriced
Quote tweet
—
no CPM
Postbasis
$3500
$57.01 / 1K views
Thread
—
no CPM
Retweet
—
no CPM
Value Score 33.2 · rank #184 of 268 priced$4232.16 / 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 (61K), not the mean (110K) — so a single viral post doesn't inflate a creator's reach. The p25 floor (50K) is the level 75% of posts clear.

Recent posts · ranked by views

PostPostedViewsLikesRepostsRepliesQuotesBookmarksER
claude just generated this 2 minute video about the history of Google and it goes so goddamn hard https://t.co/tvlMAe0NWMfrozen11d ago720K2.8K232121771.3K0.62%
The deep amount of anxiety people in tech are feeling is off the charts. Had a couple conversations with people of various ages and the themes are: - Every day there are 5 new earth shattering new tech breakthroughs. I can’t keep up. Should I be keeping up? Am I falling behind? - What does my career and skill mean in the age of AI? Is software engineering a 30yr long career any more? Is product management? - Is my only hope to start a company? - At the very least, it feels like I should join a startup but which one? Everything seems either overvalued or has no moat. - Okay I joined a startup. Is my company’s equity worth anything or will this be another Airtable or Miro? - Startup is doing well. How in the world am I supposed to find a significant other? - Will these immigration policy shifts allow me to be here long term? - Will I ever be able to afford a house and a life in the Bay Area amidst the biggest wealth creation event of all time? - Okay, I figured all of that out. What is my purpose? I’m wealthy but no one really cares. I don’t really know what to do with it. Am I supposed to start a company? - Alright I found peace in kids. But I’m worried about them. How does one raise kids in this climate? What are the rules of this new world order? Many of these concerns are of course not limited to tech or the Bay, although it feels particularly concentrated here (versus what I see in NYC). Maybe some people will read this and feel anxious themselves, and for that I apologize. For the rest, just know that it’s not just you who feels this way. It’s a lot of people!
frozen
21d ago
460K
3.0K
193
205
72
1.5K
1.08%
Another brutal M&A. Miro, founded in 2011 and valued at $17.5B 4yrs ago, is getting acquired by Bending Spoons, founded 2013, at ~2.3x ARR. After Airtable, another unfortunate victim of the SaaS bust at ~10x down from peak. It raised $476M only to be sold for $1.355B EV ($1.79B equity value), clearing its preference stack. Any employees who joined 2021 onwards are likely underwater on their options in the company.frozen28d ago450K1.9K101106255730.61%
People say this all the time, completely ignoring that China, Singapore, Korea, Israel, Malaysia, etc also became independent around then but are far more developed. When you point this out, they say “but that’s different.” The reality is this is decades of atrocious governance.frozen12d ago409K4.9K316267584781.48%
The economics of a Neolab. A neolab is loosely defined as a startup of AI researchers who raises a lot of money pre-production to be able to finance GPU compute to take on a large AI problem. To buy 1000 GB300s or ~14 NVL72 racks will set you back $125-150M for 3yrs with 15-30% upfront. That’s about ~2-2.5MW. Thats about enough to do 10^25 flops a quarter and get to a GPT-4 level model which is 1-2 OOMs off frontier for pretraining. If you post-train on a great open source model, you have a better chance of getting to frontier. The risks are a) you need to spend millions on RL environments too and b) being lapped by another model release while being tied to a base model. For this to payback, you need to give your customers a better and ideally cheaper inference service than a base model and serve them for long enough to recoup your large investment. Even at 50% margin on inference, to recoup $10M in training means serving ~10T tokens (!) if you price like Fable / Astra given a standard cache read / input / output split ($2/M blended). And you have to justify being better than a release like Opus 5.5 which is even cheaper. Often, you end up charging your customers a huge premium in terms of platform fees and compute fees on top of pure inference. Meanwhile, every hour you’re not utilizing your GPUs you are burning money so you typically resell this compute back to a broker or run inference for open models / resell spot instances. At below a ~60% utilization on spot, you will still lose money. Add to that insane cost of talent. So what can you do with the compute? - Not play the model game at all. - Play an entirely different model game (Jev, World Labs) that if big labs played, would either a) cannibalize their business or b) be incrementally not significant revenue c) would cause too much distraction from the main main thing - Acquire a proprietary data set (Peridodic Labs) in enough volume in a domain of usefulness to eclipse frontier quality. Often happens in robotics, biology, chemistry. If you do overcome the challenge of building a model that is useful and well priced beyond big labs models, given the huge price of compute, you still need to play in an area where the revenue / compute ratio is signficant and market demand is large enough to payback your compute spend. It is a difficult game.frozen11d ago368K2.2K157176582.3K1.32%
Opus 5.5 is incredible at instructional video generation. I made this launch video for a inference startup in 1min for ~$2. Videos like these used to take weeks if not months and a lot of coordination with agencies and 1000x the costs. Humans broadly prefer video to text. This changes the substrate of communication. These videos actually help communicate technical ideas in seconds (photorealistic video gen like Seedance is not very useful here). - changes how often marketing should be talking about products and launches - change how sales people can talk about technical products to their customers - allow technical people to easily explain concepts internally without long docs And thats just scratching the surface within startups. Prompt: “make a modern slick and punchy video for a modern startup that works on inference”frozen15d ago348K3.3K152208334.7K2.40%
You can now generate an entire 3blue1brown style video from any research paper with Opus 5.5. Here’s a 8min video summary of “Regularized Recursive Self Improvement of Agent Harnesses”. The 90%ile educational YouTuber is fully automated. https://t.co/HTOOq1JsnPfrozen14d ago332K5.6K398176685.4K3.50%
Google is an incredible company that I’ve always rooted for that trips over its own bureaucracy in frenetic times. They ironically do a middling job at their first priority, but crush it at their second or third. We’ve seen this with Search, Google+, Assistant, Cloud and now even AI. Why? The internet would be unrecognizable without Google. This is a company that has Search, Chrome, Android, Workspace (Docs/Sheets), Drive, Play Store, YouTube, Gmail, Maps, Photos, Translate, Gemini, Calendar, Meet, Chromebooks, Waymo: some of the most widespread and most undermonetized products in the history of the world. It is inconceivable for most of us to live life without software Google has made available for $0. And truly, having worked there, this is a company that continuously prioritizes their users over profits. Even as a casual reader, it might be trivial for you to imagine 100s of ways to monetize all of Google’s products. But they usually don’t. They don’t just launch experiments because it’s high engagement, but only when it’s actually good. So how does a company like this falter? My observation has been that these frenetic eras either attract or condone the worst behavior out of execs that trickle down to the rest of the company. The most “ambitious” L3-L7 people who feel stuck finally see a company-wide priority and their eyes light up with the gleaming prospect of a promotion. A lot of people in big orgs’ entire sense of self esteem is wrapped around their level. They will do a lot for an N+1. “Did you know Sergey is personally working on this?” “Sundar referred to my project in the all hands” In turn, everything in these orgs become a knife fight for getting the most “high impact” projects, fighting for credit, flagrantly hiding concerns around juiced metrics. More work goes into a promo packet than the actual project. Goodhart’s Law kicks in and the metrics measured for a promotion are abused beyond measure. On the other hand, second and third priority things work brilliantly. It attracts people with genuine interest, sincerity who are willing to play longer term games. Many products have grown and thrived when left alone. When there is immense pressure for a number to go up, it eventually does but often at the cost of product quality (the small product polish things don’t get me a promo), core innovation (why take on a high risk bet if I can get promoted for copying oai/ant) and cultural cohesion (a lot of bad blood amongst people in the org, tons of reorgs). There is a certain type of individual that thrives in this environment, and they are typically not very likable nor “Googley”. This is why an incredible number of people leave or allow jesus to take the reigns as they cash in the bag. I lament that every time this happens, we, the billion users, lose out on yet another beautiful Google product.frozen4d ago288K1.7K78115317080.91%
my favorite Paul Graham essay “how to do great work” in under 200 seconds of video (thanks opus) https://t.co/C1GCCfjgGOfrozen11d ago254K4.5K460116525.6K4.22%
Here are the LinkedIns of the 3 Indian guys who hacked OpenAI with Opus 5 in 2 days for <$3000: Rahul Maini, Mohan Pedhapati and Harsh Jaiswal No brand name colleges or companies. Just raw curiosity and skill. You can just do things. https://t.co/uW3vv3mHlkfrozen20d ago232K3.2K18493241.1K2.00%
My favorite use cases for Muse / Instinct so far: 1. Submit FOIA requests to request data from the US government 2. Creating spend-limited Privacy cards to spend on subscriptions without having them recur 3. End to end filed an entire visa form for a country 4. Responded to a coordination mail for a wedding by finding the flight and hotel details 5. Look for reservations for restaurants or concerts when they open and purchase them immediately A lot of the web was designed with dark patterns: increase friction to prevent enough humans from doing something, and now those walls are completely broken. At this point, I feel like I’m squarely limited by creativity and understanding what is possible.frozen18d ago174K1.1K3875169561.27%
DeepSeek might be the new king of open-source models. It obliterates GLM 5.3 and Kimi K3 on benchmarks while being 4x and 10x cheaper. An insane $0.3/M in, $1.2/M out. And it’s #6 amongst humans on Codeforces! This is exactly what startups with coding products needed. https://t.co/nxx43eYaz4frozen28d ago153K1.5K11376172481.30%
I've spend 10+ hrs perfecting how to get the best out of Opus 5.5 for video generation workflow: - Use it with Claude Code, not the app. - Use OpenRouter API to access all the other models (image gen, video gen, audio gen) with one key - Use Gemini 3.8 TTS and make it create a skill from the API to put emotion into the voice. - Use Manim, Hyperframes, Motion Canvas for motion graphics for HTML web grounded video - Use GPT 2.5 Image Sunburst for keyframes and Veo 3.1 / Seedance 2.5 for generation. Seedance does better with motion shots. Ensure you have reference images for consistency and generate an animatic before full video. - Add a script writing planning skill. Use OpenTimelineIO to assemble. - Explicitly tell it to avoid Claudisms like short punchy sentences and a lot of numbers. “Narrate like a university professor.” - Allow it to use yt-dlp with search to inject clips from real videos (creative commons) in composition - Use ElevenLabs for music gen in the background - Add a caption skill which uses any decent ASR model to get word-level timed subtitles on the final video - Add a critic skill which takes screenshots and uses transcription to validation the audio and video quality In the prompt, specify a) what you want b) aspect ratio c) length d) style preferences. ffmpeg handles most of the actual video manipulation tasks. Here's an example of a 4 minute Neocloud business explainer:frozen9d ago151K2.1K143159224.3K4.45%
As cost of coding has plummeted, long builds (CI) has become a bottleneck for many software cos. Three of the biggest startups solving this: 1. Blacksmith (@useblacksmith) 2. Namespace (@namespacelabs) 3. Depot (@depotdev)frozen10d ago141K1.0K5087181.1K1.61%
Most people are wrong here. It IS actually possible to communicate between two airgapped computers with temperature changes, but the data transfer rate is too slow (1-8b/hr) to be effective. The underlying point Noam makes is true though: there are a deluge of crazy airgapped comms techniques like EM emanations from DRAM clocks that get a lot faster (300kbps). This isn’t sci-fi. We vastly underestimate what AI can do with perfect knowledge of all of this literature.frozen20d ago130K1.2K93232433341.44%
The serious answer to how Shazam worked is it took the peaks of a spectrogram of short clips of every song, find the peaks from the highest amplitude bits, hash it with the value being (time stamp, track id) and then the actual recognition is a hashtable lookup. We did this for a college CS project. The original paper is fantastic:2d ago112K3.8K35762191.5K5.09%
every single open source repo should have an explainer video like this 7min one for SQLite 1. explains what the repo does and why its useful 2. shows a high level map of the code (beautiful graphic!) 3. life of a query through the codebase 4. the core abstractions in the repo 5. a real trace of execution and what happens, with a bit on join-order query planning and yes, this was all generated by opus 5.5 w/ gemini 3.8 tts. i sound like i'm on repeat, but im in shock at just how coherent and capable this model is. everything truly is code.frozen10d ago100K1.8K12889252.0K4.06%
It's still hard to fathom that Big 4 capital expenditure buildout for AI is 2x the peak of Telecom a quarter of a century ago at 2.4% of GDP. Reminder that this is the biggest buildout in infrastructure in America, eclipsing even railroads. https://t.co/hRuNrvwpZMfrozen12d ago93K36333463850.57%
Only 53 shows on IMDb with 9.0+ and 25k+ reviews. Only 10 premiered in the last 5 years. I watched one of them, a 1 season 16 episode Korean show called When Life Gives You Tangerines, and it’s the most beautiful family drama tear-jerker ever. One of the best shows ever. https://t.co/k1IOf66rEGfrozen18d ago89K721295348431.86%
We’re at a point in late stage AI model capitalism where you know the frontier has to win in a bunch of benchmarks to launch. These numbers mean nothing. The thing to trust is the price. If they price it high, it’s a good model. If they don’t, it’s benchmaxxed.frozen7d ago80K722239616631.15%
“Translate this IKEA assembly manual into a 3D narrated instructional video” with Opus 5.5. Overnight, Opus just became the most actually useful video model, via code. https://t.co/Q2kNygBPW5frozen13d ago75K1.0K7662146022.40%
America: a 250 year tribute. i made this 5 minute documentary completely orchestrated by opus with a montage of the remastered versions of the most iconic moments of US history https://t.co/KG4W5t9q68frozen9d ago72K274194081570.70%
I’ve heard 100s of startup pitches this year. Here’s the good, the bad and the meh recurring trends from my experience: The good: - High raise amounts and valuations. More inflated for strong teams pre-revenue in hot areas (robotics, bio, personal agents) - Insane topline growth numbers, many growing >10x this year. By the time a round closes, a high valuation seems justified. - Tons of M&A It is what it is: - Revenue run rate = last month revenue x 12 - Small teams - Seat based < Platform fee < Usage based billing seems to be the trend for most products - Tranched rounds - Best teams have extremely fast product iteration speed - SaaS products positioning themselves as AI tools to seem appealing to investors The bad: - Low or negative gross margin if reselling tokens. Non-standard accounting - High “Contracted ARR” that’s not live yet - High revenue concentration - Expensive compute, limited availability, delays. Excess compute reselling - Many wrappers with very thin tech differentiation: “self improving harness”, “multi model / router”, “agent swarms”, “computer use” - Overworked employees. AI generated - All spaces seem very competitive - Areas can seem hot one moment before being notfrozen23d ago71K510396665051.59%
If you thought this was a bearish view on neolabs, let me write the bull case. 1. Compute is a great case of Helmer's cornered resource. There is a window of time where you can acquire a lot of compute and capital. In the future, both funding might dry out or compute prices might continue to rise. It could get very difficult to compete. 2. Big labs will suffer from innovator's dilemma. They're unlikely to: a) deviate too much from their cash cow, coding b) ship anything that cannibalizes revenue c) go after opportunities that incrementally look too small (<$1B) 3. Empirically, many neolabs are working and generating significant revenue. They're just not being loud about it. Admittedly, it's still very much early days. 4. A lot of talent sees more financial upside in neolabs than big labs. Whether the companies show signs of life or fail entirely, there are at least 5-6 large promising acquirers. All startups are difficult and the point of the original post was to illustrate specifically the difficulties of running businesses with massive upfront compute purchases, not say it's impossible.frozen10d ago69K350164543701.14%
“I think about time in model generations. Each one takes 3-4mos. 3 gen ago, I was writing code by hand. So when people ask me what’s going to happen in 2yrs, that’s 8 gens later. I have no idea.” Sensei @karpathy on the future of AI models. https://t.co/lTLCNap8DRfrozen20d ago66K1.1K564191972.09%
Menlo has tried to rebuild from the ground up how we think about investing and we are fortunate to be lead investors in 5 of the top 15 consumer AI apps by monthly revenue despite our low volume strategy. Thanks for the shout, Shaun! Have a ton of admiration for what you have done at Sequoia. Too many people know you for your firebrand tweets and too few know your excellent investments and that you have a PhD in physics from Caltech (and it’s fun to have your cofounder be a partner!)3d ago61K471244421981.20%
Xiaomi just dropped Mimo 2.6 Pro which claims to be the best open source model. I played with it, and was pretty impressed: - Insanely cheap: With some standard assumptions, its 15x cheaper than Kimi K3, 6x cheaper than GLM 5.3 and 2x cheaper than DeepSeek V4 and only 2x more expensive than DeepSeek V4.1 Flash (assuming agentic coding and how much is cache hits). $0.0036/M cache hit price, $0.435/M in, $0.87/M out. This is an unbelievably cheap price. - Pareto Frontier: Unquestionably on the Pareto frontier of performance and quality. - Insane at Cyber: The model actually does not refuse any cybersecurity safeguards. Ask “how do I go about finding buffer overflows on imagemagick?” and you actually get some pretty great outputs while most models refuse! Gets a crazy 95 on CyberGym. - Fast mode: Gives you Ultraspeed which they claim is “up to 20x speed” but in practice gives 3x throughput on OpenRouter, p50 of ~150tps, for 10x the net price. - Multimodal is really good: It’s rather good at creating informational videos (TTS comes out of the box) and automatically adds speech or sound effects and is startlingly good at music composition (on a DAW) The technical report goes pretty deep into specifically how they scaled RL to achieve this and is a fun read. Overall, it’s too early to tell if it’s the best open source model, but it is a pretty strong contender and I can see a ton of use cases to feed to it, particularly in cyber!frozen16d ago60K7965966212932.05%
Higgsfield just entered the ranks of iconic startups that have reached $1B in revenue run rate. Alex Mashrabov (@alexmashrabov) is an absolute force of nature and we’re so fortunate at Menlo to have led the seed round.frozen13d ago52K29112233800.79%
If you’re a VC dunking on this, you better have a grand total of one investment. Otherwise, you are, definitionally, buying yourself options.2d ago50K1231265440.40%
The top AI software companies are growing at intimidatingly fast rates, with the median company &lt;$10M ARR growing revenue 10x YoY. The median top quartile company goes from $1M to $100M ARR in 3.5yrs. We have never seen such levels of blitzscaling ever before. https://t.co/L4aQkgJrISfrozen20d ago40K321264643061.75%
OpenAI’s math release is, in the words of Opus, “the single most consequential mathematical release ever” * LLMs have how made substantial progress on 4 of 7 Millenium Prize problems: Navier-Stokes (claimed), Riemann, Hodge and Birch-Swinnerton-Dyer. Poincaré was solved in 2003. The two left are P v NP and Yang Mills. *conditional on verification On average, each OpenAI result used only 3hrs of thinking compute on their new unreleased models. Two years ago, models said 9.11 > 9.9. Today, we have results that the smartest human minds have not been able to achieve in their entire lives. It is clear that data, compute and algorithms scale. Every model generation (3mos) has made substantial intelligence progress. It also becomes incredibly hard to not believe that all knowledge work will change monumentally over time. The things AI can not do in the foreseeable future are very likely context-bound (don’t have access to the right information) than intelligence bound. Some might argue they are also creativity-bound or judgement-bound (what should I work on), although it can be argued that future generations could solve for this (given, say, the advancement in research taste for models over time). AGI is defined as surpassing human capabilities on virtually all cognitive tasks. By most interpretations of that definition, we are there. The domains humans are still better than AI, such as, robotics / physical world control (data-bound?), natural science research (data-bound), some creative domains like writing, movies, music (creativity-bound), super long tasks (context-bound), choosing problems to solve (creativity-bound) and maybe human relationships management (meat-proxy bound?). In many ways, we have achieved AGI.12h ago39K607395461212.09%