How Much Code Is Written by AI? Depends What You Count
Anthropic, Google, GitHub and Microsoft each give a number for AI-written code. Each counts something different. Here are the claims, units, and a repo test.
We can't find the internet
Attempting to reconnect
Something went wrong!
Attempting to reconnect
Long-form essays for founders, engineering leaders, and senior ICs building at AI-native startups. Real numbers from real engagements, opinionated on what actually works. Browse all 21 topics → Earlier writing from the original Sublime Coding blog lives in the archive.
Curated reading orders by subject. Start here if you want the argument for a topic compressed into 5–6 essays rather than browsed across the full index.
What production AI engineering actually looks like in 2026: the autonomy ladder for agents, the workflow shift, telemetry, and team sizing.
A guided reading order for startup security: what to read first on SOC 2 as a revenue tool, vCISO hiring, and securing AI-native products.
Engineering leadership at startup scale: hiring from one engineer to fifteen, rituals that work at small teams, the staff-engineer interview loop.
Why language choice matters for AI systems: BEAM concurrency for agents, what Go frameworks cost you, and why Elixir is the language AI writes best.
Running Omarchy with coding agents, starting on an M1 MacBook: the install, what breaks, the keyboard, and what a security lead inherits from agents.
Anthropic, Google, GitHub and Microsoft each give a number for AI-written code. Each counts something different. Here are the claims, units, and a repo test.
Omarchy ships bash. I priced out switching to zsh against my own dotfiles: 61 of 64 aliases and my most-used function run in bash. Here is the list.
Microsoft 365 was still degraded 21 hours in. Some coverage tied it to the exit offer. Nobody has left under it yet, but the advisory shows what recovery takes.
Challenger's reports show AI leading employer-stated layoff reasons from March to July, then dropping. The numbers, and what a press release label can't say.
Most MVP cuts are safe. Five become incidents: no account recovery, no deletion, untested payments, kit secrets, open admin routes. Plus a one-hour pass.
Is `source .env` safe? Only for a file you wrote. For a cloned repo it runs whatever the file says. Here is how I load .env with direnv instead.
Plug sets cache-control on every response, so a check for a missing header never matches. How I fixed it on this site and what each page sends now.
Exit code 137 means SIGKILL, and memory is only one sender. On one Phoenix project I hit it twice: a compile that ran out of RAM, and a binary that never ran.
What I commit, merge, and keep local to run Claude Code on three machines, macOS and Linux, x86 and Apple Silicon. Plus the guard hook I forgot to track.
OpenSSH 10.6p2 shipped a workaround for sshhole on musl Linux with no PAM. Who is exposed, a five-minute Alpine check, the fix, and why there was no embargo.
A bill would pause advanced AI and ban superintelligence. What it says, what's unknown, and a vendor-continuity checklist for startups that buy models.
Claude Code's scratchpad lives in /tmp, which is RAM on my Omarchy box. 7.1 GB of worktrees froze it. The fix, what it still misses, and a checklist.