/ BLOG

Build-in-public, open notebook.

A running log of what I've built, what I've tried and dropped, and what I've looked at and decided not to touch. As much for me as for anyone else. Filter by topic, or search inline.

⌘K
agents
$ atlas callers fetchProposals
↻ re-index 1 dirty file · 20ms
✓ 3 callers · fresh · 8ms
07 JUL · 4 MIN READ

8 milliseconds. Claude Code used to take almost three minutes to find a file.

I replaced a 69,000-line code tool with 4,090 lines of Rust in a day, because a stale code graph is worse than none.

agents
#1authorcost-first
#2reviewerindependence
#3testermultimodal
#4stager1M ctx
#5pixel judgehard gate
18 JUN · 5 MIN READ

Five roles, one hard gate

Most run-phase chains optimise cost. One enforces a hard gate, because getting it wrong would ship a UI bug nobody saw.

agents
#1interviewercost-first
#2cartographercost-first
#3designerquality-first
#4panelcost-first
17 JUN · 5 MIN READ

Four roles, one frontier hire

Most jobs in a planning pipeline do not need a frontier model. Picking the right one per role saves an order of magnitude. One job earns the splurge.

agents
#1evidence
#2diagnose
#3propose
#4select
#5apply
11 JUN · 4 MIN READ

Porting RHO when you're not actually in the dark

Porting Retrospective Harness Optimisation into a harness with real gate outcomes. What the label-free recipe buys you, and what to swap.

agents
#1harness cross-check
#2delta re-verify
#3specialist fan-out
#4risk tiers
#5model routing
#6evidence bundle
#7doom-loop break
#8false-pass evals
09 JUN · 4 MIN READ

Porting Cloudflare's $1 code review into my custom agent harness

Eight techniques from Cloudflare's multi-agent code reviewer, ported into the integration-verify gate of my custom agent harness. What transferred and what I had to write down myself.