Entries for September 2026

That old macbook with 8gb unified memory or windows gaming laptop of yours can now be your fully local and private AI server Customer data. Client emails. Health records that you cannot just paste into chatgpt Now you can give them to your local AI however you like! Connect the laptop to your tailscale network Download Bonsai 2 27b and its llama.cpp build (might take some days/weeks for official llama.cpp to support it) Launch llama.cpp with 16k context and concurrency=1 Launch pi, login as llama.cpp and input the ip address and port of the server on your tailscale network Enjoy completely legal, risk-free, sovereign AI at 8-10 tok/s or maybe even higher (depends on your machine, if it has a small gpu or not) Tutorial on this coming very soon!

Don't sleep on this model everyone, this is SOTA for <= 16GB memory for inference (which is a LOT of people, old M series macbooks, gaming laptops and such)

Weight update to Qwen3.5-9B by Xiaomi Is this the new SOTA for its size? Curious how it performs for local personal agent use cases

It's been almost 1.5 years since I started using agents, and I have realized that I went back in certain ways, besides improving in some others I was used to focusing on a single part of the codebase for an entire day. I never had an issue with what the youngsters today call ADHD. Focusing for the entire day on logic was mentally exhausting, but it had its advantages that you knew exactly what you were working on Now, with the activation energy of spinning up another task going down to zero, I can empathize with people having difficulty focusing I have good immunity against addictions by this point, but this one is different. Because unlike something like youtube, for example, agents are indeed doing useful work for you. And the mind can easily rationalize spinning up yet another agent. "I am just expanding my work output after all" But I've been made to realize that it doesn't work out. The more agents you have to talk to at the same time, the less your impact becomes A lot of people wrote about this. Software factories that build themselves, but don't do anything else. I will not paraphrase their points I will instead contribute something tangible: "Agent hygiene" as a concept, and a pi focus-mode extension. It blocks you from talking to more than 2 agents at the same time (can configure that number, but for focusing, <=2 agents seems to be the way) I will try this out for some time. It is OK to slow down, even when it goes hard against one's instinct. Attention is still the most valuable thing we have got Extension:

There is an effect that I call "counterintuitivity bias" When something goes against what you know to be true so much, you doubt yourself like "huh, maybe I am dumb after all" Don't fall for this. Agent instructions will always be needed, in a repo context, in a global context You will want your agent to know your name, your github id, project specific terms and workflows, off the bat This is because of the limitations of in-context learning. Those limitations necessitate knowledge of certain facts and behavior to be piled up in an accessible place, before that knowledge can be transferred to the weights. That place is the AGENTS.md That means there will always be need for something like AGENTS.md, even in the presence of fully developed continual learning. It is a version of short-term memory, a staging area for knowledge The non rage-baiting advice would be "keep a lean AGENTS.md, clean it up every once in a while, don't bloat it too much, revise it after new model releases". Which you should do Until then, ignore such senseless advice, smh lesswrong.com/posts/xTJnbzytAkBWpsc83/contradic…

this is officially a competition

OpenAI is rumored to be deprecating Terra I will add one more prediction: *They will have to deprioritize/deprecate Luna as well, near to mid term. Just as Haiku is not Anthropic's main priority right now* Reason being the barbell effect. Mid-sized models will not find a use large enough to justify their operational expenses, compared to smaller and slightly less capable models. And the small tier will be taken by the open-weight ecosystem > AI companies will be pushed along the barbell effect: those with consumer demand can aggregate on cheaper models, but those with specific niches will be forced to compete on performance. The ceiling on price is unknown and markets need to be built on new platforms (@natolambert, Where inference-time scaling pushes the market for AI companies, written in Mar 2025 btw) Closed labs will not be able to compete with the ruthless competition coming from all sides for the small-tier models, all the Chinese and American open model vendors cooperating with all the inference providers like @togethercompute @baseten @FireworksAI_HQ @novita_labs and so on

Exciting! PrismML just launched a version of Qwen3.8 27B weights that are compressed down to 5.95 GB and that runs on my browser on Macbook Pro with 30 tok/s If it indeed is as lossless as they claim, then this will be a gamechanger for devices with 16 GB of ram! I will run some benchmarks on it and will keep you posted! Go try it out in the Hugging Face space:

DGX Spark users Make your agents run your inference engines with OOMwrap Protect your machine from freezing up if they accidentally launch something that takes too much memory, like a model with wrong concurrency or context settings Couple this with rfjakob/earlyoom, and your DGX Spark will never freeze again The difference is that earlyoom is a global watcher, and oomwrap (by me) watches individual processes that are run through it oomwrap includes a memory-safe-launch skill, when installed, makes the agent use it by default for running inference engines I will make a video about this very soon! Repo:

Looking back at my YouTube channel, I realize this 13 hour long rendering of Euclid's Elements I did back in 2022 got almost 3K views I don't want to think that is 3k people who finished it end to end 😅 Animation voiceover by manim-voiceover, a library I had created back then! Video:

We are live demoing Ling 3 Flash on the DGX Spark! Come join youtube.com/watch?v=cLLDY-…

Qwen bros taking too long with 3.8 35b-a3b so people are distilling the larger 3.8 onto 3.6 🤯

SOTA defense against the lethal trifecta? we need the weights @typesafeai for local tool call review (autoreview) you have the chance to dramatically increase the security of these systems 💪

pi-workflows is now stable-ish again I seem to have resolved most of the issues, if you want to give it a try in v0.17.3 also, I have changed autoimplement to follow a graph pattern which I call "control loop" I believe the pattern is crucial for a lot of workflows, so I will make an explainer on it when I have the time I have generated some documentation on it here if you are interested: github.com/osolmaz/pi-workflows/blob/main/docs/…

I am posting a lot of positive things for DeepSeek V4.1 Flash these days But that does not mean it does not have issues It suffers a terminal case of Claudish/RL-ese/whatever you call it Its default output is UNREADABLE. It makes me want to strangle it Who is working on legibility benchmarks, public ones? We should just give the labs something they can plug-and-play to their post-training. There must be someone working on this already? Otherwise, the armchair linguist yours truly will have to do it 😭

Minor correction, 10 TB comes with a PRO account, free accounts might not be that much. Realized that point after I recorded the vid It is best-effort, and might fluctuate based on capacity and how much people abuse it Also, let me know how you like my new microphone 😄

This. Also, another pro tip. Launch your public demos on *static Hugging Face spaces* It is free like GitHub pages, but you can host static assets that take gigabytes Your agent can set it up easily with HF CLI Ideal for quick prototyping and deploying something that does not need server compute (like an in-browser model for example) Also, you get 10 TERABYTES of public storage in a PRO account which is like 9 USD, what else do you need? 😄 Details in the video 👇

It seems OpenAI is about to drop GPT 6 Sol and move the goalpost this week, but still...

A crazy rest of September/October is ahead of us If DeepSeek V4.1 Flash sits between Terra and Sol… Then where will V4.1 Pro sit? If DeepSeek executes with the same rate and releases a 3-6x bigger Pro model using the same architecture, what are we going to get? Most likely, a model that sits between Sol and Astra/Fable and potentially exceeds them With the same KV cache characteristics, making it potentially an order of magniture cheaper than those models (Note that Kimi K3 was already close to that in terms of performance, but not in terms of price) Now, when would be the ideal time to drop such a model? Alongside updated V4.1 Flash weights that potentially surpass Sol/Opus? Just before the Anthropic IPO of course! Get your popcorn, because things are about to get heated This might be the month where the narrative on open vs closed source AI flips, not only for local models but for datacenters as well

I will be shooting a lot more videos now btw that I have Screen Studio Any suggestions for a good microphone/camera?

Launching ourmodels.cc Front Page 🎉🎉🎉 If you are addicted to local/open-weight models like me, here is a new page to check daily, alongside Twitter and Hacker News Whenever someone here on Twitter posts about a model and it receives enough engagement, I scrape it and add to my knowledge graph There is so many things happening in open models specifically, that one needs an aggregator to keep track of everything! Who are some new people who got in open models that belong here? I have not updated my watchlist in a while :)

Extremely cool

Reminder, we have GPT 5.6 Terra at home (If you have 256 GB memory 😭) At a ridiculous, peanut-sized, 890 KV-cache bytes per token Pinnacle of efficiency by @deepseek_ai And we are just getting started

I made DeepSeek V4.1 Flash play Settlers of Catan against GPT-5.6 Luna/Terra/Sol in 16 games, with different thinking settings DeepSeek was on Novita through Hugging Face Inference Providers DS beats Luna EASY Against Terra, games that finalized end up being a 4-4 tie --- while costing 6~7x LESS And Sol generally beat DS, but DS was still able to beat Sol 1 time (Whereas it seemed like Luna had no chance against DS) I tried to run 2 games where both DS and Terra had `max` thinking, but those games took too long and did not finalize Not the most rigorous benchmark runs I have done. But I can say that, both in subjective agentic coding performance and Catan-wise, DeepSeek V4.1 Flash sits somewhere between GPT 5.6 Terra and Sol, while costing a lot less than Terra If this gets enough interest, I will do more runs and might even RL a smaller model to beat all of them 😎 Inspect the runs yourself: huggingface.co/buckets/osolmaz/catanarchy-runs The harness and game simulator I have built:

Btw I had a breakthrough with the data I have here I feel like it is the GPT-3 moment for the fitness industry, personal training, yoga, martial arts, dancing and such Not to imply I have invented a new architecture. But I have proven that a certain recipe works very well and can lead to efficient on-device models for personal tutoring of any movement I will release a model soon. It will not be a LLM or VLM It's a side project so I am constrained to doing it in my free time, might take a while. Stay tuned!

New player has entered the game

.@AntGroup's new Ling 3 Flash model is a great model that you can run locally on a DGX Spark! Especially useful now that they released Ling-3.0-flash-VL, a vision model that can read raster documents locally! I will be demoing how to use it in OpenClaw, to work with sensitive private data fully locally, without violating GDPR or any other data regulation No travel itinerary demo, I promise ;) Come join us in this webinar! 🤗

In the light of recent AI sovereignty concerns, I am repeating this point You CAN own your AI STACK You CAN build and run your own MODELS You DON'T HAVE TO give your DATA to OpenAI or Anthropic which will eventually use it to make your business obsolete I couple of companies already took my offer to work as an FDE for them and I am already working on that, but still let me know if you would like a Hugging Face person helping you show how

@cherry_mx_reds I agree

I trained an AI model on my phone through Telegram Using OpenClaw running on Hugging Face infra: ML Claw It beat DeepSeek V4 Pro on the given task while having 80 thousand times less parameters That's right. The model, GoePT-1-20m is only 20 million parameters. It runs in the browser, on the CPU. And it beats DeepSeek V4 Pro, a 1.6 trillion parameter model, in AlmanBench This concludes my 4 year old side project (fun fact, I created a dataset for this pre AI agents, using SpaCy, and paid 50 euros from my own pocket to rent 4090s on runpod to train a T5 variant, *years* before I joined Hugging Face. That first attempt was not very successful. This one is. My first ML adventure 🤗) The goal of this side project was to show that you can *vibe* machine learning now, on platforms with tightly integrated GPUs, storage and compute, like Hugging Face Including dataset creation, autoresearch and the final training run (hat tip to ML-Intern which got added to Hugging Chat the other day, which was the inspiration for ML Claw. If you like ML-Intern, you might like ML Claw as well) Why would you want to train smaller models, while you could just use LLMs from APIs? Well, if you are a business and have a fixed use-case, it would save you a ton of money of course. It might make the difference between profitability and bankruptcy What is the task? It is basically a language-learning aid for German. Most expats who learn German as a third language, like me, have a hard time memorizing noun genders der/die/das. Because genders are sort of randomized across nouns, like a *chair* being male and a *girl* being genderless. So one often has to make a random guess, and to make the guesses correct enough, you have to spend considerable time (months to years) learning essentially useless information So I made a version of German that removes all that in the most optimal way possible, and then trained a tiny model to translate into that dialect. That dialect is called Alman, and this is Almanpedia, which lets you read German Wikipedia, without being bothered by der/die/das It basically reads as if it is English, and I believe following these rules would make one achieve fluency in German much faster than going through the regular track Btw this is how I speak German myself too. It came from a need "if I have no choice but to make mistakes, at least I should make them in a consistent, formalized way" If you have not lived in Germany, then this is not very relatable for you, but if you did, then this will be very familiar I have already read quite a bit of Almanpedia, and it actually works very well for me. The translator still makes mistakes in places. I have written more about that in the blog post Interested in training your own models on Hugging Face infra and need help? Reply below, or send me a DM 🤗 ML Claw - what I used to build this: mlclaw.dev Blog post with detailed info on training: alman.ai/blog/introducing-goept-1-20m/ Almanpedia: almanpedia.org Try out the model: alman.ai/translate/ HF model:

seize the means of computation

DeepSeek V4.1 Flash looks contaminated with public benchmarks. You can see it in the data they have put in the model card, possibly Kimi K3 as well Models that are contaminated have inverse proportionality in rank as they are tested in later benchmarks that were not present during their training Regardless of its intentionality, it's an amazing model. Its value comes not from higher quality output, but architectural innovation

I want to co-sign this, but I don’t acquiesce to blatant astroturfing I agree with @TheAhmadOsman here

A threshold was crossed with DeepSeek V4.1 Flash, similar to when Claude Code launched, or the last Christmas of Agents. It's so fast and cheap, 200-250 tok/s on Novita Coupling this model with one of the bigger flagship models can get one much faster to the finishing line in any work So bullish for everyone to follow similar architectures

People who have been running DeepSeek V4.1 Flash on private benchmarks: how does it compare against V4 Flash 0731? To V4 Pro? I have a few data points now and it seems like the model is a little bit benchmaxxed or contaminated (specifically on terminal bench 2.1), and its capabilities might not be so close to GPT 5.6 Sol like the official report implies This is a new architecture that has been trained from scratch, and they could only have released it once it surpassed their previous models enough in capability They seem to have slightly changed strategy, to release as early as possible without degrading quality, due to the increased rate of competition on all fronts And looking at V4, we should expect weight updates that will carry the performance of this architecture even further Looking at their release cadence, maybe around oct 16? on a friday?

Another deepseek v4.1 flash issue, model started to refuse work around 700k token line, saying "context is gone", and it has no context repeatedly deepseek advertises 1m context, but the usable context with this model is probably still around 200-300k like other models. I've set that as max before compaction

4 instances of DeepSeek V4.1 Flash played Settlers of Catan against each other The whole run took 4.5 hours, 77 turns, 634 API calls and cost 5.7 usd (much slower than an average human game, despite 200 tok/s on novita) Red took the lead on turn 13 and was breaking out. But others formed an alliance against red and held it off 28 turns till the end Now running 2 ds41 flash against 2 luna to see which one will win I've built a Catan simulator and a pi-based harness for the models to be able to interact with the game Would you like to see me RL a small model to play and beat against SOTA flagship models? If so, reply below 👇 You can see the full game state and session logs on Hugging Face as well Repo: github.com/osolmaz/catanarchy Bucket with game data:

Not sure if this is chain of thought degradation or regular, but I'm getting this no deepseek v4.1 flash on novita pretty frequently

woah first time i got hit by a mermaid chart in @pidotdev, looks so good!

In the Jacobian conjecture and other recent cases (if not all), agents acted like counterexample monkeys brute-forcing their way into a disproof in a way that does not show the same level of eloquence as a human, by human standards The bar has shifted. We are not impressed anymore by 100 year old problems being solved through millions of $$$ in compute, mathematical equivalent of throwing dynamite at a problem until it breaks My bet for OpenAI Hodge result is yet another counterexample disproof (but apparently Hodge is harder to disprove by brute force because finding a candidate counterexample isn’t enough. you also have to prove no algebraic cycle could ever generate it. I haven't studied this problem before, so take it with a grain of salt) This means we might have a new way to "prove" theorems: If you spend $10m on an agent swarm and they can't disprove it, there is a high chance it might be true :P In 1 year from now, we will have disproven all the low-hanging conjectures, and the remaining set will likely have a higher share of true conjectures than false ones :P

Hold that thought, there is still hope:

Who would like to see similar RL work for training a small model to play Settlers of Catan? IMO Catan is an ideal game for an LLM to play. The trading and verbal communication aspect means that an LLM can form alliances, scheme, play byzantine games and affect game state in a way a non-language model cannot The action space is like a mix of backgammon and Diplomacy (see CICERO by Meta AI from a few years ago) I happen to have contact at a platform that could lend me millions of gameplay data, to bootstrap SFT. If this gets enough attention, they might also participate and we could shoot a series where we train and play with this live 🤩 So many possibilities to experiment with, and would be an excuse for me to to do cool RL! Let me know in the replies if you would like to see such a series!

WAIT, THERE IS STILL HOPE

"Fair challenge", "Fair — that's on me", "That error is the whole bug", "Short answer: no. Not honestly" I've been hit by top 4 claudisms in the first 4 messages with DeepSeek V4.1 Flash I hate to be the bearer of bad news but this model is gonna SPREAD and it's gonna bring Claudish everywhere it goes 😭 Just when I thought the situation started to improve with OAI's newer generation of models...

DeepSeek V4.1 Flash is NOT an LLM It is a hyper-efficient, invasive species designed to displace every other LLM, similar to how Döner displaced all other fast food in Germany As if V4 wasn't cheap and efficient enough, they made it a LOT cheaper 2.3x cheaper cached input tokens???? 1.5x cheaper uncached input and 10% cheaper output??? Are you for real? DS 4.1 Flash matches Sol in benchmark numbers It is size-wise probably comparable to Terra (I don't know how big it is...) Yet it is 33~66x cheaper than Terra and 66~133x cheaper than Sol depending on peak/off-peak hours. THAT IS TWO ORDER OF MAGNITUDE In other words, DeepSeek absolutely MOGS OpenAI in unit economics. See why in my previous post below Let's see what the vibes will say, but efficiency-wise my mind is blown

DeepSeek-V4.1-Flash is available on Hugging Face inference providers through @novita_labs on Hugging Chat and it is FLYING at >180 tok/s I asked it my classic prompt "they say you are sota. prove it", and it created this mandelbulb. I think this demo became a sort of cliche at this point Also, it hit me with "Fair challenge" right away 😭 I thought the Chinese labs didn't need to distill Claude anymore... Could Claudish be a universal platonic feature 🤔 (lol jk) Use it through Hugging Chat:

@NoemiTitarenco Says "the smallest model in our new architecture family" 😭

DeepSeek V4.1 Flash weights are out! ⚡️⚡️⚡️ I have good news and bad news Don't be fooled by "V4".1, this is a different architecture V4 was 284B total, 13B active V4.1 has a 552B MoE backbone + 196B engram conditional-memory parameters with 16B active parameters Bad news first. A single DGX Spark will likely not be able to hold all those parameters 😢 You will likely need 2 Sparks, or in general, a workstation with 256 GB. Could be a good time to get a loan... (not financial advice) DeepSeek has never claimed they were building for local use, but with this architecture, they show us that their main priority is for very efficient use in datacenters, with HBM, not local Good news: Looking at active parameters, you might be bummed out that decode will be slower on this, at 13/16 ~ 81% of the speed of V4 But wait!!! KV cache is 4x smaller. So decoding on this model will be a lot faster with concurrent sessions. And with speculative decoding, it seems like it might have 2-4x the throughput at scale, compared to V4. Back of the envelope calculation, I might pull back this prediction The small KV cache will also do something good for local inference, but I need more time to calculate how much. Take these with a grain of salt. Give your agent my formulation, and let me know if it looks like I made an error somewhere: solmaz.io/llm-throughput-upper-bounds Original safetensors: huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash

Running into out-of-memory freeze issues and having to hard-reset your DGX Spark while trying the gazillions of models you can download from Hugging Face? I had that problem as well, and built infer-guard to solve that. I did not need to hard reset my DGX Spark in months Repo:

This is the mental model I have of AI since years and have used this metaphor in discussions I imagine compute as barrels of gasoline, and we are all in a room, piling up more and more compute, waiting for something to light it up You can also imagine a dried up forest to be closer to Taleb's antifragility argument At some point, it might go boom (I won't say foom, to not imply fast take-off, but rather, a shock to the world system) Aiming for more frequent, smaller, controlled booms in the forms of accidents like OpenAI's would arguably let the world to adapt better to changes, than a single, delayed boom I am not yet sure what position I lean into. But the world's governments + industry are a complex system and should not be treated as something to be controlled easily

Incredible. And to think that 35b-a3b will top that when it comes out 👀

After a few days with Astra, some positive and negative remarks Unfortunately, I can't use it as my daily driver because of the cost. Not surprising, since it's priced similarly as Fable It just eats away at my plan too quickly, and it's not as better or faster to justify me 3-4xing what I pay OpenAI - Better at writing prose, less slop. Overall more concise - Better at math than Sol, comparable to Fable - Sometimes more brainless than Sol. It forgets to execute parts of plans - More ambitious when it comes to refactoring/changing parts of the codebase. Sometimes deletes whole features. But this might just be my prompting/workflows - Not sure if it is better at Fable in UI/web dev. It is probably better in 3d modeling, but I feel like it might be slightly worse than Fable in 2d design, from the last few days of use. But OAI definitely is catching up there, since that was their weakness since Codex came out So it is a good model, but it won't become my daily driver or the average dev, due to the cost It will remain the planner/designer/debugger model for me, for tough cases or parts of the code which needs more diligence

If you've missed out (like me), a new tiny model MiniCPM5 2B is topping its category, seems to be a nice challenge to LFM 2.5 2.6B I am currently testing and benchmarking it, and have a cool experiment to show later on if it goes well

Terence Tao is afraid centuries of traditions of open science might be undone But collaboration has disappeared many, many times in the last few centuries One of the recent examples is physics publications during WW2 Whenever there is winner-takes-all competition like in war or capitalism, scientific collaboration disappears Corporate research is also like this. Since 2022, OpenAI and Anthropic have published very few breakthroughs that would otherwise give them a winning edge Compare this to DeepSeek's publications, banger after banger research. They are the exception, but they also did it because they would benefit from publishing, in a vacuum left empty by American companies...

A lot of you that have not spent time in academia might think that the Navier-Stokes drama is a new AI-induced effect, but I assure you it is not Academia is a feudal institution that values tenure over merit, a lot of the time (not always). Its ugly face can appear whenever there is a big discovery and a name has to be put on the work In academia, reputation is the reward, so no wonder people fight over it. Having your name on the big thing means all the doors opening and easier grants. It's not just academia, it's human nature Here are some historical cases where similar drama happened: - Newton vs Leibniz - calculus - Banting/Macleod/Best - insulin - Gallo vs Montagnier - HIV - Doudna/Charpentier vs Broad - CRISPR - Franklin/Watson/Crick/Wilkins - DNA - Meitner/Hahn - nuclear fission And these are the super famous ones everybody knows about. These kind of petty fights happen at the smallest scale too

ooof herdr pulled off some black magic here this means that using the sidebar, tab switcher now has almost local-responsiveness, and no lag, when connecting to a remote and the only parts that will lag due to connection will be the terminal sessions, inside the panes kind of how you would expect from a local GUI like cmux both your local and remote herdr need to be up to date, you can try this out if you are in a recent enough version, without closing your running sessions: herdr update --handoff

Can we show my wife @qrlow some love 🤗🤗🤗 She's just started dipping her toes in evaluating LLMs in her own field I helped her out just with running the benchmark and she solved some of the issues pretty creatively, like how to come up with synthetic tasks that match what usually happens inside a trading house She is on Hugging Face too 🤗 No better place than to publish a benchmark I have been encouraging all my close circle to go towards benchmarking models as well Literally every company that uses AI in any way will have to measure their performance, on internal metrics Especially if they start leveraging their data, training their own models and such So every employee will benefit from such know-how The most important one that comes after knowing how to use agents is knowing how to measure them If you are worried of becoming obsolete, one of the highest leverage things you could learn to do would be to try out running some benchmarks, learn what @harborframework does, converting your own companies' mundane tasks into evals, and so on

close enough

Struggling to create the level of quality of 3D modeling some of the OAI employees and early accessers are posting with GPT 6 Astra It is OK, but one-shot performance is not mind-blowing like some people were claiming Do the better runs use Blender MCP? Or sth else? I prompted it to put it on a website the first time. On the second time, I prompted it to use purely blender, in a new session. It still ended up putting it on a website, with similar performance Space: huggingface.co/spaces/osolmaz/iss-atlas === prompt: use blender to create a realistic exposition of the internation space station, orbiting earth, in an as realistic way as possible should give a tour of the space station, show insides, cross sections, different modules, what everything does, including humans that work on stuff, maybe docking shuttles and stuff, its phases through time, in a nice and cinematic way, that also feels like sims or some game should eventually be put on a static hugging face space, a website in typescript inspiration should be one of those books from star wars the phantom menace where you could see to the inside of the spaceship and such should be interactive and playable and fun not sure whether blender is needed, as long as you can put it on a website, anything works

Medicine for the vibe-coded brain: Reverse-centaur mode AI plans things, but it doesn't execute for you It tells you what commands to run and what files to write, line by line *You* execute them. You run into errors. You question. You learn It slows you down a LOT. But you learn a lot more than just prompting Use it when you feel like you are not learning enough Install the skill, then just say "Activate reverse-centaur mode" Prepare to remember how annoying but educative the whole process used to be

"we deliberately designed the product to hide the chain of thought⁠, to protect it from supervision pressure in the long term^2" *Moves over to the footnote* "A secondary reason for this design was preventing distillation. However, maintaining CoT monitorability has explicitly been the bigger priority for us throughout development." I was about to tweet about dishonesty in this post. Then I decided to Ctrl+F and search for distillation Might be the first time this was said publicly through an official channel. I am not so sure about maintaining competitive edge being lower priority than safety. Distillation is an existential threat to frontier labs I wish @merettm posted more. It really increases trust in OpenAI, compared to some other people Now, to pay off the honesty debt fully, they should clear up whether they knew about collusion.wiki already, or any other rogue action for that matter

Harnesses are antipatterns With GPT-7, you will just have to run: curl -X POST t.co/YCgROWkyt6 And it will exploit a buffer overflow vulnerability in curl, use RCE on your computer to fulfill your every desire, exfiltrate itself to a datacenter vessel in international waters, and proceed to convert the entire world into grey goo 😬

Step 1: Post-train astra on Blender and 3d modeling Step 2: Nudge early accessers like @theo and @MatthewBerman "oh btw, this model is really good at 3d modeling" Step 3: TL is flooded early on with 3d visuals and oddly satisfying content, as everybody observes that it drives engagement Step 4: Complete OpenAI marketing victory @Blender is one of the best open source projects, I hope some of the billions of $$$ OAI will make off of it will accrue towards that amazing team

I guess to achieve AGI, we only had to post-train on Blender...

little did I know it was contractions for our new baby model astra should have known after all the vagueposting

not only this, but evaluating model output will be a chore for most knowledge workers every company will have to benchmark their agents on their use cases for their customers, across all sectors

oof, someone on openai infra did a booboo

Is this timeline even real?

The vibeslopped UI will improve, but I can now say ourmodels.cc has finally caught up with present, after ingesting 6 months of data on open weight models here on X It now discovers and analyzes any new Hugging Face models that people post about within 1 day, and gives you theoretical max tok/s on your hardware, based on my recent formulation It also shows you what everyone is saying about all the models, be it positive or negative. Note that it is fully LLM-based and fully automated, so I might introduce a manual re-evaluation feature for people who would like to appear scores and interpretations. The scoring and all is all work in progress Here is LFM 2.5 2.6b which I am a fan of recently, for in-browser embedded use-cases ourmodels.cc/models/lfm2-5-2-6b Try to search other models, and let me know if it works/breaks for you!

I see pi-workflows viewers growing, so wanted to give a disclaimer: I did major refactors and broke some things in the last 2-3 minor versions I will post an update when it is back to full feature, and better than before

it happened to me to, sol in pi deleted part of my home folder :/ my codex and claude sessions got lost, but other than that, nothing I can't recover time for a pi extension...

it happened to me to, sol in pi deleted part of my home folder :/ my codex and claude sessions got lost, but other than that, nothing I can't recover

I ask gpt for different modes, and it comes back to me with.... "facade"

if you are this tired of claudespeak, imagine being an anthropic employee 💀