Work in progress: 181,232 of 181,233 sites checked so far, more every week · updated 2 Nov 2026Missing a blog? Tell us

Top all-time · #10279

D

Divyansh Agrawal

Personal blog · divagr.com/blogaiprogrammingscience

Divyansh Agrawal's personal blog at divagr.com covers ai, programming and science. Since 2026, 7 different Hacker News readers have linked it 7 times, most often “Qwenputer: A functioning, deterministic RISC-V CPU inside Qwen3.5”. It is still writing: the last post was on 2026-10-30, and you can follow it by feed.

Visit the blog
7mentions?
7HN people?
2026–2026first – last linked?

Too few sampled comments (4) to say how often HN recommends it.?

Why HN links it

Commenters link the Qwenputer post, which describes a deterministic RISC-V CPU built inside Qwen3.5 weights. Other comments link the author's resume and Factorio simulator project. None judge the writing.

Summarised by Claude from a sample of 4 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.

Research notes and experiments on machine learning systems, reinforcement learning, evaluations, and language models.

Most-linked posts?

  1. Qwenputer: A functioning, deterministic RISC-V CPU inside Qwen3.5
    4 people · 4 mentions · first linked 2026-10-28
  2. /resume.pdf?v=20260922
    2 people · 2 mentions · first linked 2026-10-29

What Hacker News says?

  1. While we were having this discussion, someone went ahead and did this: https://divagr.com/blog/qwenputer

    Comment linking /blog/qwenputer

    on “Three new discoveries in mathematics, computer science, and physics”
  2. Submitted Qwenputer: A functioning, deterministic RISC-V CPU inside Qwen3.5 as a story · 702 points

    HN discussion
  3. Who needs DOOM when you have Factorio? https://divagr.com/resume.pdf?v=20260922 Projects FactorioRL | GitHub | Reinforcement Learning, Simulation, Sim-to-Game Transfer – Built a tick-exact Factorio simulator at 947k decisions/sec on 16 CPU threads (∼3,600× the real engine) with trace-level parity, driving RL training at 26,800 steps/sec on an RTX 4060. – Transferred simulator-trained policies to the real game: agent…

    Comment linking /resume.pdf?v=20260922

    on “Qwenputer: A functioning, deterministic RISC-V CPU inside Qwen3.5”
  4. Submitted Qwenputer: A functioning, deterministic RISC-V CPU inside Qwen3.5 as a story · 7 points

    HN discussion
  5. https://divagr.com/resume.pdf?v=20260922

    Comment linking /resume.pdf?v=20260922

    on “Qwenputer: A functioning, deterministic RISC-V CPU inside Qwen3.5”
  6. Submitted Building oversight tools into a model's weights as a story · 2 points

    HN discussion
  7. A couple thoughts on the process here… First off, Opus 5.5 built working weights and code with only two prompts: 1. Read this article and any relevant source code: https://divagr.com/blog/qwenputer . Then explore if you could emulate the chip and ram in a Nintendo Entertainment System (as people do today with FPGAs in MISTRs) using this technique. If so, what size model would you have to build and if you could use s…

    Comment linking /blog/qwenputer

    on “Show HN: A working Nintendo in ~1GB of Qwen weights”

Search all of HN for divagr.com