Product Decisions

Articles about the product side of hard technology — how a capability gets defined, what counts as evidence that it works, and what happens to the product manager’s job when specifications become probabilistic instead of deterministic.

Personal AI agents compared: Claude Duet vs Dot vs Gemini Spark vs Grok Bot vs Muse (October 2026)

One table comparing eight personal AI agents: hosted Claude Duet, Dot, Gemini Spark, Grok Bot, Instinct and Muse, plus open-source, self-hosted Hermes Agent and OpenClaw. Every cell links to an official page; anything the vendor doesn’t publish says “Not disclosed”.

An agent's shippable capability is set at the review gate, not the generation gate

22% of merged PRs are agent-written, ~0.6% agent-reviewed. That gap is the spec, not a model lag.

Personal agents split into two markets: productivity sells ROI, secretary sells time

Productivity agents sell proven ROI to a narrow paying base; secretary agents sell back time to everyone else, monetised by transactions, not subscriptions.

Eval as spec: grade the agent's side effects, not its success story

Judge agents by executable end state and pass^N, not self-reports; the eval harness is the product spec, and verification depth is the constraint.

Personal agents: the entry point buys the habit, the interaction buys the trust

The entry point buys the habit; the interaction — what the agent hands back and who verifies it — buys the trust.