Skip to content
View adamentwistle's full-sized avatar
:shipit:
managing the herd
:shipit:
managing the herd
7 followers · 3 following

Achievements

Achievement: Pair ExtraordinaireAchievement: Pull Sharkx2Achievement: StarstruckAchievement: YOLO

Achievements

Achievement: Pair ExtraordinaireAchievement: Pull Sharkx2Achievement: StarstruckAchievement: YOLO

Highlights

  • Pro

Block or report adamentwistle

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
adamentwistle/README.md

a split-flap departures board. adam entwistle, applied ai engineer, quicknode. agentic systems you can measure and trust. each row is a project, with its github stars as the flight number and its last push as the remark. the last row is agi, delayed. a ticker reads: please do not leave agents unattended. unattended agents will be merged to main.

remarks come from the last push. a github action rebuilds the board every morning, based on herdr-departures. agi is running late.

i'm an applied ai engineer at quicknode, where i build agentic systems you can measure and trust, on top of web3 infrastructure.

outside work i build tools for people running a lot of coding agents at once, and post the patterns and the dead-ends on x. most of it started as something i needed myself. i build for the way my own brain works, so the work survives my attention span. it turns out that's good design for anyone working with agents.

my thesis: ai agents are the next billion users of the internet, and the tooling layer is being built right now.

now boarding

fable-skills benchmark, mean score bare then with skills: haiku 4.5 44 to 53, sonnet 5 74 to 80, opus 5 90 to 95. fable 5.1 running bare scores 94.

fable-skills · 39 engineering-discipline skills, each one a habit written down. with them loaded, opus 5 matched fable 5.1 running bare. no deps, mit.

orchi running three claude code workers on a project and asking for a decision in a side panel

orchi

herdr set up for running many claude code agents at once, with a layer that plans the work and hands it out. needs a mac. 		a pixel-art office with one desk per claude code session, each character animated by what its session is doing

claude code mods

every claude code session as a desk in a pixel-art office, animated by what it's doing. plus two more mods.
a treemap of a repo with named agents sitting on the files they are reading and writing

herdr territory

your repo as a live map, with your coding agents moving around it. you can watch two of them walk into the same file. 		a claude code wrapped share card showing tokens processed, activity and api-equivalent cost

claude code wrapped

wrapped, but for your claude code usage. run it, get a share card, find out if your plan is a bargain or a problem.
  • legwork · a project queue for claude code that survives you walking away. /wrap mints the next prompt while context is hot; /pickup briefs you back in thirty seconds.
  • agent-eng-kit · small stdlib-only tools for agent engineering: a prompt-injection gauntlet, an mcp record/replay proxy, a session profiler, a context packer, an mcp doctor. no deps, no keys, mit.
  • claude-code-tools · zero-config doc linters for claude code projects.

flight log

github stats: total stars, contributions in the last twelve months, public repos, current day streak and a heatmap of the last year

arrivals  06 Nov 2026

ground crew

python go typescript react next.js tailwind css fastapi wails claude api claude code mcp github actions langfuse inspect ai playwright


adamentwistle.com   x: @aae_on_x   linkedin

Pinned Loading

  1. fable-skills fable-skills Public

    39 engineering-discipline skills with a blind-graded benchmark: opus 5 loaded with them matched fable 5.1 running bare. no deps, mit.

    Python 98 11

  2. claude-code-wrapped claude-code-wrapped Public

    wrapped, but for your claude code usage. run it, get a share card, find out if your plan is a bargain or a problem.

    Python 3

  3. agent-eng-kit agent-eng-kit Public

    small stdlib-only tools for agent engineering: a prompt-injection gauntlet, an mcp record/replay proxy, a session profiler, a context packer, an mcp doctor. no deps, no keys, mit.

    Python 3

  4. claude-code-tools claude-code-tools Public

    zero-config doc linters for claude code projects.

    Python 2

  5. orchi-build orchi-build Public

    Herdr's look and panels, and an orchestration layer for Claude Code agents

    JavaScript 16

  6. legwork legwork Public

    a project queue for claude code that survives you walking away. /wrap mints the next prompt while context is hot; /pickup briefs you back in thirty seconds.

    Python 1