← The Independent 100: a study

Raymond Weitekamp

#94 in The Builders on Main, the curator’s rank and section @raw_works Relevance to our work: 5 of 5

Bracketed tags are evidence labels: [Verified] means we read the page, file or API response ourselves, [Reported] means a source says so and we did not check the underlying thing, and [Inference] is our own reasoning.

Credit

The Independent 100 list was curated by dan (@irl_danB) and published on 2026-09-24 at independent.prose.md (announcement). The ranks and section names are his, and number 100 is his open wild-card slot. This study is ours and is not endorsed by him or by anyone on the list.

What they publicly do

Founder of polySpectra and a resident at Prime Intellect [Reported: X profile x.com/raw_works]; he builds recursive-language-model (RLM) coding agents and publishes benchmark runs on raw.works [Verified: raw.works]. He is #3 of 11 contributors to OpenProse (MIT) and says he developed the Skill Safety Data Sheet standard [Verified: GitHub contributors API; raw.works/engineering-skills].

Public work

  • ypi repo MIT last activity 2026-06-22 · 391 GitHub stars on 2026-09-24

    A recursive coding agent built on Pi. A single rlm_query tool spawns child agents with the same prompt and tools, isolated in jj workspaces, with five guardrails: depth, call count, dollar budget, timeout and PATH scrubbing. It includes a make doctor and parity tests.

  • dirpack repo MIT last activity 2026-08-04 · 14 GitHub stars on 2026-09-24

    A Rust CLI that builds token-budgeted directory indexes with tree-sitter signatures for coding agents. It ships a Claude Code SessionStart hook, eval baselines on DSPy/build123d, and a .beads tracker.

  • OpenProse repo MIT last activity 2026-09-15 · 1,748 GitHub stars on 2026-09-24

    A Markdown 'language' executed inside the agent session: responsibilities declare Requires/Maintains contracts and are wired by the model. His PR #62 added deterministic skill declarations, and he wrote the auto-pocock example.

  • Skill Safety Data Sheet (SSDS) other no license stated

    An open standard (v0.2.0) that rates agent skills on Host/Data/Actuation/Content hazards (0-4). Hazard flags are derived deterministically from a capability manifest, including prompt-injection surfaces, then mapped to containment levels and required approvals. It has a Zod schema with sha256 artifact pinning.

  • LongCoT / StudyBench harness experiments blog no license stated last activity 2026-06-20

    Reports Codex CLI with gpt-5.5 xhigh at 1446/1995 (72.5%) on full LongCoT, against the paper's claim that math and chemistry stay near zero. StudyBench library/lab ablations are published with their environments.

  • rlm-cli repo no license file last activity 2026-06-18 · 91 GitHub stars on 2026-09-24

    A CLI wrapper for RLM with directory-as-context and JSON-first output.

Licenses of GitHub repositories are from the authenticated GitHub license API on 2026-09-25, one row per repository in evidence/licenses.jsonl (“no license file” means the API found none; “license file present, not recognised by GitHub” means GitHub could not classify it). Other licenses are as the source states. Check the license file before reusing anything.

Relevance to our work

5 of 5 lenses: agents, orchestration, evals, rust, security, research-method

Several MIT artifacts sit squarely in our lanes: a bounded recursive agent (ypi), a Rust budgeted repo index with a hook integration (dirpack), and a contract-style agent workflow language (OpenProse). SSDS is a ready-made hazard rubric for vetting skills. His 'it's the harness, not the model' benchmark method matches franken-research's evidence grading.

Scored 0 to 5 against the study’s relevance rubric: 5 means public, openly licensed work squarely in our lanes that we could study or adopt now, and 1 means little connection to our work. The score is one reviewer’s judgement of how much this public work bears on our own work. It is not a ranking of people and says nothing about the person or the quality of their work.

What we would study next

  1. Map SSDS HDAC scoring and hazard flags onto a skill intake and security scan (skillsafetydatasheet.org). It gives a deterministic hazard score and required approvals for each skill, which complements static skill security scans. Effort: medium.
  2. Benchmark dirpack against other repo-orientation tools on one Rust repo (github.com/rawwerks/dirpack). A Rust, MIT, fixed-token-budget index with a SessionStart hook; measure the token savings on real repositories such as Jeffrey's crates. Effort: small.
  3. Study ypi's guardrails (depth, budget, call cap, jj isolation) for bounded subagent recursion (github.com/rawwerks/ypi). Directly comparable to how OMP and NTM bound fan-out; the parity tests are a model for harness compatibility gates. Effort: medium.

Before reusing anything: ypi, dirpack, VibeCAD and OpenProse are MIT per the GitHub license API (2026-09-25). rlm-cli and studybench-lab-and-library have no LICENSE file (all rights reserved by default). rawwerks/recursive-coding-agents has a proprietary 'All rights reserved, no license is granted' LICENSE: do not reuse it. No SSDS license was found on skillsafetydatasheet.org. dirpack's repo has age-encrypted internal-private-docs.

Sources

Overall confidence in this record: High. Sources were read on 2026-09-24 and every link was checked on 2026-09-25. The full record is line 94 of people.jsonl.

Limits

Not verified: The license of the SSDS standard text and schema, and the repo it lives in. His role at the 'AIE' organization named on his profile. The LongCoT numbers are self-reported and were not reproduced.

This record comes from a first-pass survey of public material by an AI agent session on 2026-09-24, edited for publication on 2026-09-25. Nothing was installed or run for it. Roles, stars and licenses change; see the study’s limits.

Request a correction

Is something about this record wrong or out of date, or would you rather not be covered? Open a correction (a GitHub issue form with this page’s address filled in). A request to remove a record is honoured without argument; accepted corrections are recorded in the changelog with the date.

Help fix it

Each opens a GitHub issue form with this page’s address filled in. You need a GitHub account.

Made by Joshua Nowak (ZestStream) with AI coding agents. Not affiliated with or funded by Jeffrey Emanuel; Joshua uses his public tools and is a paying subscriber to jeffreys-skills.md.

Source on GitHub MIT licensed