Ask a model to write a lock-free ring buffer in an empty directory and you will have working code in fifteen seconds. Ask the same model to ship a routine feature inside a three-year-old repository and you will often watch it spend twenty turns reading the wrong files, editing a dead helper, and handing back a pull request that takes a senior engineer an hour to untangle.
Tools for Agentic Executive Functioning
Models can write almost any isolated function on the first try. Where software teams still stall is the work around the code: keeping a plan intact across fifty files, noticing when a repository has drifted into something neither humans nor models can navigate, and knowing what actually shipped.
When that layer is missing, buying more tokens doesn’t move the roadmap faster—senior engineers just spend half their week untangling diffs.
Products
Fractal OS
fractal123.com
Work breakdown structures, critical-path scheduling, and sub-agent execution over a shared project.xml and chat.xml document.
use.sh
use.sh
Postgres, Redis, Mongo, and IRC instances provisioned from a single command so neither humans nor agents wait on environment setup.
Invariant
invariant123.com
Continuous SQL invariants and metrics queried directly from production Postgres to catch silent data drift as code ships faster.
devtee
devtee.com
Opt-in session recording that turns real engineering workflows and keystrokes into reproducible agent evaluations.
srclint
CLI · Python
Traditional linters check whether a single file parses cleanly. srclint and cq check whether an entire repository is structured cleanly enough that an engineer or coding agent can find the right module on the first read.
cq
CLI · Python
Finds oversized files, import cycles, and naming drift. Static checks for structural issues across a repository.
tokdump
CLI · Rust / Python
Like hexdump, but for tokens. Prints byte offsets, token IDs, and token boundaries in a three-column hexdump(1) layout.
tc
CLI · Rust / Python
Like wc, but for tokens. Counts tokens in files or stdin using o200k_base.
metadata-enforcer
CLI · Python
Django-style schemas for Markdown frontmatter. Validates and fixes Markdown frontmatter against schemas written in Python.
Writing
Beyond Chat Scrollback: Storing Plans and Agent Runs in One XML Tree
Why plans in Jira and execution in ephemeral chat windows drift apart, and how project.xml keeps tasks, dependencies, and sub-agent transcripts in one place.
Counting What the Model Pays For: tc and tokdump
A single lockfile or inline SVG can quietly burn 40% of a context window on every turn. Two small Unix tools make o200k_base token weight visible in local shells and CI.
What does ‘redacted’ mean? Designing a sentinel for computer-use telemetry
Why the obvious redaction markers fail for computer-use data, and what a better one looks like.