Top all-time · #1936
Prime Intellect
Company blog · primeintellect.ai/blogaiprogramming
Prime Intellect is the company blog at primeintellect.ai, writing about ai and programming. Since 2024, 37 different Hacker News readers have linked it 44 times, most often “Recursive Language Models: the paradigm of 2026”. No feed was found, so read it on the site itself.
Too few sampled comments (13) to say how often HN recommends it.?
Why HN links it
HN readers link Prime Intellect's posts and research on decentralised and RL model training, synthetic verified datasets, and the NanoGPT speedrun benchmark.
Summarised by Claude from a sample of 13 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Train, deploy, and continuously improve your own models on an integrated compute, training, inference, and sandbox stack.
Most-linked posts?
- Recursive Language Models: the paradigm of 20265 people · 5 mentions · first linked 2026-01-01
- Measuring Autonomous AI Research4 people · 4 mentions · first linked 2026-08-16
- Lab: The Full-Stack Platform for Training Your Own Models3 people · 3 mentions · first linked 2026-02-11
- Intellect-3: A 100B+ MoE trained with large-scale RL3 people · 3 mentions · first linked 2025-11-27
- NanoGPT Speedrun Frontier3 people · 3 mentions · first linked 2026-08-22
- /2 people · 2 mentions · first linked 2025-07-09
- INTELLECT–1: Launching the First Decentralized Training of a 10B Parameter Model2 people · 2 mentions · first linked 2024-10-11
- Synthetic-1: Scaling Distributed Synthetic Data Generation for VerifiedReasoning2 people · 2 mentions · first linked 2025-02-12
- Uncovering a universal offline sandbox escape2 people · 2 mentions · first linked 2026-08-25
What Hacker News says?
- nl · 2026-08-24
> Codex is more token efficient This is very true, especially vs Opus 5. > tends to get better results than Fable with less need for extreme token burning shenanigans like 18 hours of subagents. The strength here was less the code quality and rather the long horizon task tracking. This was a very large task - I was chatting with the maintainers and we estimated 4 to 6 months work over multiple phases for human codin…
Comment linking /research/nanogpt-speedrun
on “Anthropic's best AI model struggles to attract users as cheaper tools thrive” Submitted Intellect-2 Release: The First 32B Model Trained Through Globally Distributed RL as a story · 201 points
HN discussionThere are references about escaping offline sandbox.. I dont know about shootings!! https://www.primeintellect.ai/blog/universal-offline-sandbox...
Comment linking /blog/universal-offline-sandbox-escape
on “There are no "rogue" AI agents”Submitted NanoGPT Speedrun Frontier as a story · 140 points
HN discussionRunpod is worth a look for these on demand workloads https://www.runpod.io/pricing I use a lot for ffmpeg workloads. Found this a few days ago which might be neat for finding cheaper https://www.primeintellect.ai/ No affiliation with either
Comment linking /
on “Smollm3: Smol, multilingual, long-context reasoner LLM”Submitted INTELLECT–1: Launching the First Decentralized Training of a 10B Parameter Model as a story · 111 points
HN discussionSynthetic data is already widely used to do training in the programming and mathematics domains where automated verification is possible. Here is an example of an open source verified reasoning synthetic dataset https://www.primeintellect.ai/blog/synthetic-1
Comment linking /blog/synthetic-1
on “The Bitter Lesson Is Misunderstood”Submitted Rewriting Prime Agent in Rust as a story · 64 points
HN discussion>The idea implies that eventually crowd-sourcing an open AI is probably technically feasible It's already technically feasible: https://www.primeintellect.ai/blog/intellect-2
Comment linking /blog/intellect-2
on “Huawei releases an open weight model trained on Huawei Ascend GPUs”Submitted Prime Agent: A self-improving RLM agent as a story · 18 points
HN discussionThere are a couple: https://ambient.xyz/ https://www.primeintellect.ai/
Comment linking /
on “The upcoming GPT-3 moment for RL”Submitted The First Globally Trained 10B Parameter Model as a story · 8 points
HN discussionWhat's going on with sol here? The note says it spends a lot of time waiting, did it just not effectively use time (i.e. something like parallel runs) so it's graph ends up being stretched in the time axis? I'm also seeing notes like on Opus 5 saying it was a run with a older serial version of program.md, so the graphs aren't complete apples-to-apples comparisons? Edit: the blog seems to address these https://www.pr…
Comment linking /blog/measuring-autonomous-research
on “NanoGPT Speedrun Frontier”Submitted Peer-to-Peer Decentralized Inference as a story · 4 points
HN discussionCool post! I'm a co-author on a recent blog post from METR about the NanoGPT speed-run here [1]. I think it'd be of interest to anyone who enjoyed the original post. Appreciate the good beefy runs and spend here, it's a (from my experience) not super easy to do! (Also: just to label this comment clearly: it's written hastily from a car, and based on lighter skim of the original blog post [2] than would be ideal. Ple…
Comment linking /blog/measuring-autonomous-research
on “NanoGPT Speedrun Frontier”Submitted Intellect-3: A 100B+ MoE trained with large-scale RL as a story · 3 points
HN discussionI think that's explained here: https://www.primeintellect.ai/blog/measuring-autonomous-rese... Basically they do 8 runs trying to optimize to under 3.28 loss in the fewest training steps possible under time/token constraint. I dunno why 18 * 8 != 153 (it's 144)
Comment linking /blog/measuring-autonomous-research
on “NanoGPT Speedrun Frontier”Submitted Environments Hub: A Community Hub to Scale RL to Open AGI as a story · 3 points
HN discussionTry Fable and make it write a ton of tests. I generally don't trust benchmarks, as they don't match my experience, but this one[1] very much does. [1] https://www.primeintellect.ai/research/nanogpt-speedrun
Comment linking /research/nanogpt-speedrun
on “Fast and Hard Code”Submitted Moondream is joining Prime Intellect as a story · 2 points
HN discussionLimits but doesn't prohibit. See https://www.primeintellect.ai/blog/intellect-3 - still useful and can scale enormously. Takes a particular shape and relies heavily on RL, but still big.
Comment linking /blog/intellect-3
on “Coding Agents Could Make Free Software Matter Again”Submitted Uncovering a universal offline sandbox escape as a story · 2 points
HN discussionI mean it as in, train a model across different clusters instead of a centralized cluster. It's been shown that it's possible to train 10B models this way. If more research effort was put into this, that would be great https://www.primeintellect.ai/blog/intellect-1
Comment linking /blog/intellect-1
on “GPT‑NL: a sovereign language model for the Netherlands”Submitted Systematic Reward Hacking and Prime Sprints as a story · 2 points
HN discussionThis blog explains context folding ie tool calling and recursion - https://www.primeintellect.ai/blog/rlm
Comment linking /blog/rlm
on “Ask HN: What is best way to provide continuous context to models?”Submitted Intellect-3: A 100B+ MoE trained with large-scale RL as a story · 2 points
HN discussionSubmitted PCCL: Low-level, fault-tolerant communication library for decentralized training as a story · 2 points
HN discussionSubmitted Decentralized Training in the Inference-Time-Compute Paradigm as a story · 2 points
HN discussionSubmitted Prime Inference: Fast, Reliable Serving for Frontier Open Models as a story · 1 points
HN discussionSubmitted GLM-5.2 RL weight transfer in 4 seconds using NIXL and ModelExpress as a story · 1 points
HN discussion
