AI Update, October 2026: Notable Shifts for People Who Build Products
Trends we have been seeing: CLI agents are maturing, subscriptions come in more tiers, local models are worth a look, and verification tooling is moving to the center.
4 min read
Local-first AI tooling: what stays on your machine and what doesn't
A plain description of data custody when you use AI agents: what lives locally, what providers see, why bring-your-own CLI helps, and the honest limits.
5 min read
Prompt → contract → verification: how we make agent work checkable
Write the done-criterion first, decide what a machine can check versus what needs a human, and keep evidence receipts so agent work can be trusted.
4 min read
Multi-agent orchestration: the mistakes that cost us weeks
Agents that say done without evidence, retry loops, file ownership clashes, and our rules: two failures means stop, measure output not activity.
5 min read
Agent CLIs compared from daily use: Claude Code, Codex, Gemini CLI, OpenCode
An opinionated, practice-based look at four terminal coding agents: what each does well, where each tends to break, and how we split work between them.
4 min read
Why we run coding agents on subscriptions, not API keys
Predictable cost, quota as a budget, and the cases where pay-per-token API access is still the right call for coding agents.
5 min read