Product & Craft · Week of September 27, 2026
Agent factories scale while compute scarcity, AI pricing, trust, and retention tighten
- AI Software Factories Hit Production Scale as Review and Accountability LagWarp ships 2,000 pull requests a month, and the first one for a new request can be ready in as little as 35 minutes. A ticket typed into Slack moves through Linear and GitHub, gets tested, and comes back as a merged PR. …
- Agent Tool Use Turns Compute Scarcity From GPUs Toward CPUs and Energy348.41 Wh. That is what one Reflexion agent on Llama-3.1-Instruct 70B spent answering a single HotpotQA question, 136.5 times the energy of the same model doing one-shot inference.
- Microsoft Consolidates Copilot as AI Pricing Moves to Credits and Flexible BundlesA Copilot Credit costs one penny.
- Muse and Voice Agents Expose Trust Gap Between UX and Model QualityPeter Yang ran the same Japan flight itinerary through Grok Bot and Meta's Muse. Muse came back with a price more than $1,000 higher.
- Retention Pressure Shifts From Magic Moments to Activation Debt and FrictionActivation was healthy. Onboarding completion was high. The magic moment event fired for most new users in their first session. Three months later, more than half of those users were gone — no single dramatic churn momen…
Threads we are following
- Who will own the review and accountability bottleneck as AI software factories scale?
- Whether CPU-first scheduling and OS-level agent isolation become standard before the agent-era CPU shortage eases.
- Whether CIOs can build governance that distinguishes valuable agent runs from expensive polling loops before the credit meter resets.
- Whether voice agents can make tool provenance and confirmation legible in audio without sacrificing conversational fluency.
- Whether longer AI-app trials can offset variable inference costs without eroding conversion.
One issue a week. Free.
sift. reads hundreds of sources so you don't have to, and sends one issue you can actually finish. No daily noise. Unsubscribe anytime.