https://wato.org/blog/ A commercial telco-security product shipped with fleets of coding agents: real incidents, the rules they produced, and numbers that survived a hostile audit of git history. en https://wato.org/blog/a-reviewer-that-has-to-prove-it/ https://wato.org/blog/a-reviewer-that-has-to-prove-it/ Fri, 30 Oct 2026 06:00:00 +0000 From the record № 1. In six days, one person and a fleet of coding agents built Proven, an open-source pull-request reviewer that reports only what it can prove. What the record says about the week, including the parts that did not go to plan. https://wato.org/blog/five-things-a-telco-product-taught-a-fleet/ https://wato.org/blog/five-things-a-telco-product-taught-a-fleet/ Wed, 23 Sep 2026 06:00:00 +0000 Five posts, one product, one summer. The fleet has moved on to a new job — the lessons have not. The closing card of the agent-fleets series: one lesson per post, with the link to the evidence, and what the series becomes next. Post 6 of the agent-fleets series. https://wato.org/blog/fail-open-fail-closed/ https://wato.org/blog/fail-open-fail-closed/ Wed, 02 Sep 2026 07:00:00 +0000 A security product sits between the customer and what they are trying to do, so when it breaks it has to break in a direction. We wrote the direction down in April; in June a phone that changed networks 26 times in four minutes showed us the one layer the sentence had not reached — and taught us something about first fixes. Post 5 of the agent-fleets series. https://wato.org/blog/first-commit-was-a-sales-strategy/ https://wato.org/blog/first-commit-was-a-sales-strategy/ Wed, 26 Aug 2026 07:00:00 +0000 The oldest commit in the requirements repo is a 700-line sales strategy and a 923-line deep-link spec — six days before the first mock existed. Where those 1,662 lines went. Post 4 of the agent-fleets series. https://wato.org/blog/agents-couldnt-agree-on-a-tech-stack/ https://wato.org/blog/agents-couldnt-agree-on-a-tech-stack/ Tue, 04 Aug 2026 07:00:00 +0000 Every fresh agent session is a new hire — day-one memory, year-five confidence. The five-and-a-half-hour evening that fixed it, and what a requirements repo is actually for. Post 3 of the agent-fleets series. https://wato.org/blog/thirty-minute-test-window/ https://wato.org/blog/thirty-minute-test-window/ Tue, 21 Jul 2026 07:00:00 +0000 A payment sandbox open thirty minutes a day, physical SIM cards, a team across three time zones — and the mock that turned the window into a confirmation ritual. Post 2 of the agent-fleets series. https://wato.org/blog/seven-mocks-in-ten-days/ https://wato.org/blog/seven-mocks-in-ten-days/ Mon, 06 Jul 2026 07:00:00 +0000 The operator-side systems you can't run on a laptop, the extraction that bent the cost curve, and what git corrected about my memory. Post 1 of the agent-fleets series.