A developer's quiet notebook — practical notes from experiment to production
The 2026 AI Index from Stanford HAI still carries plenty of data, but the insights land with less weight than last year. What the report says, what it misses, and why an annual cadence has fallen out of step with how 2026 actually moves.
AI products are fundamentally different from traditional software. This post distills 12 counterintuitive lessons from 20 builders, the CC/CD framework for AI development lifecycles, and a 3-phase eval system — all sourced from Lenny Rachitsky's newsletter.
Self-Tuning Loop 4 steps (Generate → Capture → Analyze → Evolve) extracted as a universal module. Supabase DDL, diff capture utilities, analysis/evolution prompts, email/blog examples, GitHub reference implementation.
The actual production system implementing Self-Tuning Loop from Part 1. Data collection (35 sources), AI curation, Telegram input pipeline, weekly auto-review, and Syncthing-based zero-deploy prompt evolution.
What surfaces on the second read of Anthropic's Advisor Tool. This isn't new architecture — it's a temporary fix shaped by 2026 pricing. A pattern that disappears once Opus prices drop, and eleven other papers from the same period are quietly moving the same way. The anchor of the series.
How do you organize the sliced agents. Four structures, paired with one skeptical paper that argues 'LLM swarms aren't really swarms.' The conclusion lands where it usually does — structure choice gets dragged along by pricing. Part 2 of the series.
The five ways 2026 papers slice an agent, side by side. Two things stand out by the end: Role, Skill, and Judge are different names for the same concept, and the time-axis literature is nearly empty. Part 1 of the series.
Operational review of WICHI's backend on Railway and Supabase, covering real-world incidents like connection pooling and migration conflicts, plus criteria for infrastructure migration.
Analysis of the design philosophy and hierarchical dependencies of the four GEO Score layers (Inclusion, Prominence, Quality, Stability), and their implications for decision-making.
Introducing Self-Tuning Loop: capture implicit feedback from human edit diffs, analyze patterns periodically, and auto-evolve prompt guidelines. Includes academic gap analysis (DSPy, TextGrad, POHF).
Documentation of WICHI's 9-Bucket query framework. Defines 3 Zones and 9 Buckets based on brand presence to measure AI's organic recommendations effectively.
Analysis of multi-engine architecture design principles that leverage response variance as signals, featuring parallel collection structures and scalability via the adapter pattern.
A guide to zero-cost minimum monitoring using Sentry and Betterstack, showing how an 80-minute pre-launch investment reduced error response time from 13 hours to 1 hour.
Documenting the psychological mechanisms and viral loop strategies behind three sharing elements designed for MMU's PLG growth: score cards, badges, and custom checklist links.
