← The Independent 100: a study

dex

#87 in The Builders on Main, the curator’s rank and section @dexhorthy 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

Runs HumanLayer [Verified: humanlayer.com], which now sells a 'multiplayer coding agent workspace' [Verified: humanlayer.com]. Author of 12-Factor Agents and of 'Advanced Context Engineering for Coding Agents' (frequent intentional compaction, research-plan-implement) [Verified: humanlayer repos]. Runs the 'AI That Works' podcast [Verified: github.com/ai-that-works/ai-that-works].

Public work

  • 12-Factor Agents repo license file present, not recognised by GitHub last activity 2025-09-21 · 26,380 GitHub stars on 2026-09-24

    Twelve principles for production LLM software: own your prompts, own your context window, tools are structured outputs, unify execution and business state, launch/pause/resume, contact humans with tool calls, own control flow, compact errors, small focused agents, stateless reducer.

  • Advanced Context Engineering for Coding Agents (ace-fca.md, wsff.md, SlopCodeBench write-ups) blog no license file last activity 2026-08-04 · 2,659 GitHub stars on 2026-09-24

    Frequent intentional compaction: keep context utilization around 40-60%, research then plan then implement, human review at the plan; later essays on why software factories fail and on benchmarking Opus 5 with SlopCodeBench (quality over evolving checkpoints).

  • humanlayer/skills repo MIT last activity 2026-09-17 · 4,483 GitHub stars on 2026-09-24

    Claude Code skills: design-control-loop (set point, sensor, controller, actuator, dampener, scheduled PR loop), build-iterated-agentic-loop, improve-claude-md (<important if> blocks), visual-pr, show-me.

  • shannon repo MIT last activity 2026-05-13 · 249 GitHub stars on 2026-09-24

    CLI/SDK that runs a real interactive claude session in tmux, sends the prompt, tails the on-disk transcript and emits stream-json, instead of using claude -p.

  • fold repo MIT last activity 2026-09-24 · 75 GitHub stars on 2026-09-24

    Effect-native, provider-agnostic agent core with log-based state, hooks, subagents, auto-compaction; picks edit tools per model (write/edit for Claude, apply_patch for Codex).

  • AI That Works podcast no license file last activity 2026-09-21 · 1,961 GitHub stars on 2026-09-24

    Weekly session repo with code from each episode.

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, research-method, writing

MIT skills (design-control-loop) and shannon are adoptable now; 12-Factor Agents and ACE-FCA are the most-cited practitioner doctrine for context management; SlopCodeBench write-ups point at a benchmark for code quality over time.

Deep dive:

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. ace-fca.md: frequent intentional compaction and research/plan/implement with review at the plan (github.com/humanlayer/advanced-context-engineering-for-coding-agents/blob/main/ace-fca.md). Direct counterpart to compaction and handoff practice built on post-compact hooks and beads; its 40-60% context target is a testable rule for long agent sessions. Effort: small.
  2. design-control-loop skill (MIT) (github.com/humanlayer/skills/blob/main/plugins/design-control-loop/skills/design-control-loop/SKILL.md). Formalizes a scheduled agent loop as sensor, controller, actuator and a dampener regression gate, each runnable locally first; the dampener gate applies to any scheduled agent loop. Effort: small.
  3. SlopCodeBench benchmarking notes and the paper (github.com/humanlayer/advanced-context-engineering-for-coding-agents/blob/main/benchmarking-opus-5-on-slop-code-bench.md). Measures maintainability as requirements evolve across checkpoints; a candidate external oracle for grading agent code quality over time, not only pass rates. Effort: medium.

Before reusing anything: 12-Factor Agents: code Apache-2.0, text and images CC BY-SA 4.0, so quoting or adapting its prose triggers ShareAlike [Verified: README license section]. humanlayer/humanlayer is Apache-2.0 but its README says the code is 'pretty much all deprecated'. The ACE repo and ai-that-works have no license file (study only). humanlayer/skills, shannon and fold are MIT. shannon automates the interactive Claude Code CLI; check Anthropic's terms before running it at scale [Inference].

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 87 of people.jsonl.

Limits

Not verified: The BAML results in ace-fca.md (300k LOC Rust, 35k LOC shipped in about 7 hours) are his claims. HumanLayer's current product was not installed or tested. No recent X posts were pulled for him.

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