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
CTO of HumanLayer, which sells a multiplayer coding-agent IDE and cloud (sessions, plan artifacts, local and cloud daemons) [Verified: github.com/K-Mistele; humanlayer.com]. He writes about MCP security, context engineering and vLLM tool calling, and was earlier an offensive-security writer [Verified: blacklight.sh/feed.xml].
- X profile: x.com/0xblacklight
- GitHub: github.com/K-Mistele
- Site: blacklight.sh
- Blog: blacklight.sh/blog
- Also: humanlayer.com
- Also: github.com/humanlayer
Public work
- fold repo MIT last activity 2026-09-24 · 75 GitHub stars on 2026-09-24
An Effect-native, provider-agnostic, isomorphic agent core with an optional coding agent, CLI and TUI (
foldcode). It has a log-based event state, hooks, subagents, skills, auto-compaction and per-model tool selection (edit for Claude, apply_patch for Codex). Kyle is the only listed contributor (212 commits). - humanlayer/skills repo MIT last activity 2026-09-17 · 4,483 GitHub stars on 2026-09-24
Claude Code skill plugins: show-me, visual-pr, design-control-loop, build-iterated-agentic-loop, improve-claude-md. Kyle is #2 of 3 contributors (4 commits).
- humanlayer/humanlayer repo license file present, not recognised by GitHub last activity 2026-06-19 · 11,605 GitHub stars on 2026-09-24
The original HumanLayer human-in-the-loop SDK and CodeLayer; the README now says the code is mostly deprecated. Kyle is #2 of 28 contributors (451 commits).
- MCP Deep Dive: the Great, the Broken, and the Downright Dangerous blog no license stated last activity 2025-07-09
A field report on MCP's design, auth problems and security flaws.
- Implementing OpenAI-compatible tool calling in vLLM blog no license stated last activity 2024-10-30
Transcript of his talk after landing vLLM's OpenAI-compatible tool calling and tool streaming for open models.
- opensearch repo MIT last activity 2025-07-01 · 30 GitHub stars on 2026-09-24
An open-source deep-research agent, with a build write-up on his blog.
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
4 of 5 lenses: agents, orchestration, mcp, security, local-models
fold is an MIT agent harness core that can be benchmarked against OMP's design (event log, hooks, subagents, compaction, per-model tools). His MCP-security and vLLM tool-calling writing bears on MCP server security and local-model tool use. HumanLayer's product is closed, and fold is young (75 stars).
Deep dive:
- dex, Kyle Mistele and HumanLayer (12-Factor Agents, ACE-FCA, skills, shannon, fold) (written 2026-09-25)
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
- Read fold-core's event-log state, hook and subagent model and compare it with OMP (github.com/humanlayer/fold). A clean, typed harness core to benchmark other harness designs against. Effort: medium.
- Read 'MCP Deep Dive' for its security findings and check MCP servers against them (blacklight.sh/blog/mcp-deep-dive-great-broken-dangerous). Direct input to prompt-injection and MCP auth hardening. Effort: small.
- Check the design-control-loop skill in humanlayer/skills (github.com/humanlayer/skills/tree/main/plugins/design-control-loop). Its control-loop taxonomy may map onto cron-driven agent loops. Effort: small.
Before reusing anything: fold, humanlayer/skills and opensearch are MIT. humanlayer/humanlayer and 12-factor-agents carry an Apache-2.0 text, but the API reports NOASSERTION and humanlayer/humanlayer is deprecated. K-Mistele/swarm is GPL-3.0 (copyleft). advanced-context-engineering-for-coding-agents has no license. The HumanLayer IDE and cloud are proprietary.
Sources
- blacklight.sh/feed.xml
- github.com/K-Mistele
- github.com/humanlayer/fold
- github.com/humanlayer/skills
- github.com/humanlayer/humanlayer
- humanlayer.com
- x.com/0xblacklight/status/2102924412010070215
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 97 of people.jsonl.
Limits
Not verified: Which parts of HumanLayer's current closed product he personally built. Authorship of the advanced-context-engineering benchmark write-ups. His vLLM contribution beyond his own talk transcript (not checked in the vLLM git history).
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.