Blog

RSS feed →

Two approvals instead of four

What changed in great_cto: the pipeline stopped needing a courier, and one check worth thirty seconds before you read the rest

Two weeks of teaching my orchestrator to stop lying to me

86 commits under one slogan: the system is allowed to not know — it is not allowed to not know confidently. A green-tests-broken-UI story, a benchmark that caught its own judge cheating, and why an empty list is the most dangerous answer an agent system can give.

GreatCTO now builds compliance-reviewed software for regulated industries — 5 new verticals, 20 new products

The catalog grew from 10 US industries to 15, and 40 products to 60 — but the news isn't the count, it's the kind. The five new verticals are all regulated: allied health, dental, insurance, accounting & tax, and law firms. Describe a product in one of them and the pipeline attaches the matching compliance reviewer — HIPAA, NAIC, SOX, or the bar rules — automatically.

Stage zero: a product-owner agent now runs before the architect — and the one human gate moved with it

The pipeline used to start at the architect, so the first decision you approved was HOW to build. But the most expensive mistakes are made before that — in WHAT to build. So we added a product-owner agent that frames the problem, runs a 4-model idea debate, and hands the architect a validated brief. The gate moved to it.

June under the hood: the board becomes a pult, prompts evolve behind a holdout gate, logs shrink 99.5%

Approve a gate → an agent spawns and streams live. Plus: a self-improvement loop with anti-overfit gating, $0 context compression, scope-pinned task briefs, and Fable 5 support.

The operator console: where the autopilot's work waits for a signature

Durable runs, an inbox for licensed humans, a signature ceremony for irreversible writes, and an Ops tab with a dead-letter queue. WCAG 2.2 AA, axe-core: 0 violations.

We pivoted: GreatCTO is now AI autopilots for business

From 6 to 25 verticals in one week — each runs on live connectors, and the runtime physically refuses to fire an irreversible action without a human signature.

great_cto: what's new — three features and the move to Opus 4.8

A discovery pipeline, a quota-warning hook, a digital-health pack — and a no-drama model upgrade.

Everyone is squeezing context. We stopped putting everything in one context.

87.7% tokens per pipeline run. Not by squeezing — by splitting.

great_cto v2.17 - no more tambourine dance

One install, everything works. Companion plugins, jurisdiction-aware agents, 16 new board features.

AI Agents Work While You Sleep — Now They Can Wake You Up

Added email alerts and browser push to the board. Because 'check the terminal every 5 minutes' is not a workflow.

Real cost breakdown: 10 packs, $0.60 LLM bill, $42K saved per regulated feature

Per-feature, per-MVP, per-quarter numbers. Hardware ratios, runway math, and the honest places where the savings stop.

What $1.4M of compliance work looks like in 14 hours – ten packs, ten regulated industries

Startups have often reached out to me with the same problem: their team could ship a regulated feature in days, but the compliance setup around it took weeks and tens of thousands of dollars.

The MTTR -94% claim, with receipts

47 paired P0 incidents across 12 repositories. 4 honest misses. Full methodology + how to replicate the measurement in your own repo.

Three days of code, six weeks of compliance — the math behind why

Not a complaint about lawyers. A breakdown of where the six weeks actually go, and which parts of it are mechanical.

How GreatCTO chooses which compliance pack to attach

Regex vs LLM-based archetype detection, the false-positive count, and why I keep rejecting the obvious fix.

Why your agent system fails: missing gates, not missing intelligence

The bottleneck in agentic SDLC isn't model quality — it's process governance. Here's the state machine that closes the gap.

How I designed the SDLC state machine for agentic coding

Eight stages, two human gates, four memory layers. Why this exact shape, and what I tried that didn't work.

First real shipped feature with this stack — receipts

One run, one feature, from prompt to merged PR. Time, cost, and gate-by-gate breakdown — no marketing math.