Pinned Tweet
NJDine.com To help restaurants hit hard by the current Covid-19 restrictions, your friends at Dinecube Solutions Inc built a free mobile app named NJDine. Please visit njdine.com @GovMurphy @FairLawnPD @njdotcom @northjersey
This is a new open decision model with 31B parameters built on Gemma 4.
• The model is #1 on the Jev Decision Index leaderboard
• 2x faster than Jev
• 20% cheaper than Jev
• Built on Gemma 4 but outperforms Gemini Decision
You can run this locally!
• Input: text and images
• Output: structured answers + confidence scores
You can use this model to ask:
• yes/no questions
• multiple-choice questions
• rate different options
Picking the next tool shouldn’t take a paragraph.
We built Karpo Decision. A family of six open-weight models, from 0.8B to 31B.Give it the context, a question, and the options. It returns a probability for each option in one forward pass.
Karpo Decision 31B
🏆 73.9 on Decision Index, tied for #1 on the Nov 12 board
⚡ 94 ms median per request on a single RTX PRO 6000
📊 87.2 on JevBench Capability vs. 78.5 for Google’s Gemini Decision 1
🤗 Hugging Face
huggingface.co/MachinePulseA…
📝 Technical write-up
karpo.ai/blog/karpo-decision…
Here’s where I’d try it first: Search. Recommendations. Tool choice. Model routing. Browser actions. Point it at the slowest decision in your stack and tell us what breaks.
Matt retweeted
Texas Governor-elect Gina Hinojosa:
“I stand with every single border sheriff in support of safe communities.
Public safety is not a partisan issue.”

ALT Screenshot on a grey card of a Threads post by Gina Hinojosa (@ginafortexas, verified), posted 18 minutes before the screenshot. The post reads: 'I stand with every single border sheriff in support of safe communities. Public safety is not a partisan issue.' Below the text is a group photo of Hinojosa standing with about twenty border sheriffs, many in cowboy hats, in front of an adobe building. The post shows 706 likes, 17 replies and 34 reposts.
Matt retweeted
🚨 BREAKING: Texas Governor-Elect Gina Hinojosa to ensure all Texas children have quality healthcare.
Matt retweeted
did an incredible fun and spicy chat with @HarryStebbings !!! he even PROMISED to dye his hair pink 🩷
So Diogo Almeida is one of the most fascinating CEOs I have ever interviewed.
born in the Philippines
2008: wins bronze for the Philippines at the International Mathematical Olympiad
2013: starts his career as a software engineer at Amazon
2017: joins Google Brain, working on cutting-edge deep learning research
2018: leaves Google Brain and takes roughly 18 months off to figure out what makes him happy
travels the world, experiments with improv, art and meditation
becomes a top-30 League of Legends player
discovers that none of it makes him particularly happy
2020: joins OpenAI when it's still a relatively obscure research laboratory
helps develop reinforcement learning from human feedback (RLHF), the technique that transforms how people interact with AI
2022: co-authors the landmark InstructGPT paper
the work becomes foundational to ChatGPT
November 2022: ChatGPT launches and becomes one of the fastest-growing consumer products in history
2023: contributes to GPT-4, one of the most important model releases in AI history
but starts questioning the fundamental assumptions behind the technology he helped build
2024: leaves OpenAI while ChatGPT is exploding in popularity
people think he's crazy for walking away
co-founds TypeSafe AI with Erik Gafni and Sasha Sheng
spends two years in stealth developing an entirely new class of AI models
September 15, 2026: launches Jev, TypeSafe's first System One model
Company makes $100M in revenue in the first 7 days of launch
raises a reported $40M seed round led by DCVC
October 2026: a reported $870M Series A at a $7.5B valuation, backed by a16z
argues that the next breakthrough in AI will come from making intelligence cheap, reliable and autonomous
My notes from my discussion with @CompleteSkeptic below:
1. Why Jevons' Paradox Does Not Cut It. We Are Going to 10x That
Diogo expects a "mega Jevons paradox", with usage that dwarfs the token consumption of LLMs and AI running in the background everywhere. He expects people to stop seeing AI as something that "slops crap at them" and start seeing it as something that makes life better.
2. Why SaaS Incumbents Are Better Positioned Than OpenAI and Anthropic in Many Application Spaces
SaaS companies know their customers' workflows and have the scale to fund the upfront engineering that real automation requires. Diogo argues their edge comes from being closest to the problems, not from distribution. Teams inside products like TurboTax know where to improve the interface, while OpenAI and Anthropic are "still living in their ivory tower."
3. Why We Do Not Hire From OpenAI
Diogo sees getting into OpenAI as an anti-signal and says its researchers love algorithms and scaling but hate data. Data is a low-status topic at most labs, and the few people focused on the right task would want to pursue their own. That leaves researchers obsessed with algorithms, which he likens to throwing money in the furnace.
4. How the AI Bubble Could Burst and Where the Demand Shock Could Be
Part of today's AI demand is driven by FOMO, from people believing they are getting good results. Diogo finds many buyers miscalibrated and overestimating what AI can do. Once they face the realities of its capabilities, demand could fall quickly.
5. What Metric Did ChatGPT Optimize for in the Early Days?
Before Claude Code, ChatGPT was optimized for monthly active users rather than the best chatbot experience, which is how sycophancy emerged. Companies are now optimizing for the MAU equivalent in coding agents. Without a dedicated model, Diogo doubts they will produce a comparable one soon.
6. Why the Talent Market Is Not Crazy for Us and We Don't Pay for Mercenaries
TypeSafe pays people good amounts but not the highest, which filters out hired guns. Diogo says Anthropic is suffering from mercenaries because it pays more than anyone else, and an entire market now exists to prepare candidates for Anthropic culture interviews.
7. We Wanted Sonya Huang More Than We Wanted Sequoia
Diogo did not want Sequoia's money and brought the firm in because of Sonya. He values her particular shape of brilliance, which offers a perspective no other advisor gives him. One TypeSafe value is to make great decisions slowly, and he sees her as a valuable thought partner for that.
(links in comments)
Matt retweeted
Just ten days since Pete Buttigieg was reelected as Transportation Secretary and already flights grounded nationwide
Matt retweeted
Finally!!!!
#OM Trailer ❤️
To the sweetest fans, thank you for being so enthused and patient!!!
Much love to you all ❤️
youtu.be/aGSOpb_6kv8
Matt retweeted
I have a major result to share.
I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China.
This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model.
Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves.
I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems.
I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used.
It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20.
These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs.
I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo….
There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.
Matt retweeted
I thought people already thought/knew this?
I have a major result to share.
I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China.
This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model.
Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves.
I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems.
I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used.
It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20.
These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs.
I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo….
There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.
Matt retweeted
Justice delayed but not denied
Whistleblower: I kept coming after Ken Paxton dallasnews.com/opinion/comme… via @dallasnews
Matt retweeted
Someone that lacks this kind of integrity in his personal life will absolutely cut corners in his professional life. This is exactly why intelligence services target people who are serial cheaters.
A third Jensen affair (with another NVIDIA employee) just dropped. reuters.com/technology/forme…
Matt retweeted
Improving model accuracy doesn't always require additional training.
We increased GLiDE's Decision Index score by 1.85 points using its existing weights.
Common approaches to improving model performance include scaling parameter count, adding training data, or increasing inference-time compute through generated reasoning tokens, but these can all come with significant costs.
Another way is recurrent looping, which increases a model's internal computation by running certain layers multiple times before producing an output.
Some middle layers of transformers operate in a relatively shared representation space, and repeating them can potentially refine the model's internal representations without introducing new parameters or increasing model size.
Our initial experiments focused on training GLiDE to process information through recurrent loops, but these methods added training costs and sometimes even degraded model performance.
Instead, we found that we could apply recurrent loops directly to the already-trained model. By selectively repeating two groups of middle layers, with partial updates to limit changes to the hidden states, GLiDE was able to score 73.37 on the @huggingface Decision Index, surpassing our previously top-ranked model.
We break down the details of our experiments, the results, and key takeaways in our latest blog.
Blog: fastino.ai/blog/recurrent-lo…
Matt retweeted
This is a message for @realDonaldTrump.
Donald, this is on the table. Guarantee healthcare for every American, from cradle to grave.
Matt retweeted
JUST IN: 🇷🇺🇮🇷 President Putin discusses plans with Iran to build an oil pipeline between Russia and Iran.







































