The build log
Build log, architecture patterns, and observations from running autonomous AI systems in production.
An event-pricing prototype looked simple until it met hybrid checkout, existing feature gates, proration, account-state rules, and legacy billing. The hard part was making the product decisions explicit across those systems.
The abuse filter caught an attacker. It was us. Our own redirect chain was tripping the speed flag on real admin accounts.
A login abuse filter blacklisted a power user for clicking too fast; the IIS logs proved the fast clicks were our own redirect chain, not an attack.
A long mobile-rendering investigation blamed the platform before inspection showed a responsive reorder had left the banner frame without the child that supplied its usable height.
An internal overview page accumulated more than 80 blocking queries per load. The important fix was to reconsider its page contract and defer nonessential sections, not merely tune one query.
Manual setup can leave an unseen human in an agent workflow. A lifecycle-ownership audit for every dependency, starting with the browser, reveals which steps are not yet safe to run unattended.
A mobile-button complaint looked like a style ticket, but the same investigation found a boolean error in the layout-save path that could remove modules without a visible explanation.
A hotfix day produced a pre-edit branch guard that blocked protected-branch writes in this setup, alongside a docs rebuild that gave AI-assisted work a more usable reference surface.
A merge conflict in an agent skill file delayed a hotfix and exposed an important change-control lesson: files that shape repeatable automated workflows need intentional review.
A request called for a low-discoverability event signup path. An existing admin linking flow already supplied the needed context and downstream handling, so extending it avoided a parallel implementation.
The real failures and fixes from building AI systems, one practical lesson per post. Get the next one in your inbox.