00an open lab
What's actually running underneath.
Experiments from every layer of the stack. What was built, what broke, and the numbers that came out of it.
- layers
- 04
- experiments
- 05
- flagship
- stackhealth.dev
- rev
- 2026-10-09
fig. 00 — the stack, exploded viewclick a part
01interface
Where people touch the system.
Front ends, bots, the surfaces that hide every layer below them.
parts list
01.aweb front endsreserved
01.bchat surfacesreserved
01.cdashboardsreserved
fig. 01 — interface, wireframedashed = not yet shipped
02agents / llm
Models wired to tools, and where the wiring fails quietly.
Field notes from running agents outside a demo.
fig. 02 — agent loop, one request● = message in flight
03code
The part everyone ships and nobody measures the same way twice.
part 03.1 · flagship · live
StackHealth
An open code-health benchmark with a formula you can read.
- inputs
- any public github repo
- outputs
- security · quality · hygiene · community
- formula
- open, versioned
- cost
- free
fig. 03 — stackhealth, repo to scoredial values illustrative
04infra
Containers, tunnels, the box under the desk.
Plenty running here. None of it written up yet.
parts list
04.ahostrunning
04.bshared db / cacherunning
04.ctunnel · per-app tokensrunning
04.dobservabilityrunning
fig. 04 — infra, request pathdashed = undocumented