~/blog · filtered by category

product.

9 entries filed under this signal. ← back to all entries

[blog/category/product] · reverse chronological$ ls -la --filter=product

2026-07-17productPAELLADOC now has inference benchmarksA local module for measuring the route your model actually takes — not repeating somebody else's leaderboard.#inference-benchmarks#ai-models#model-selection#ai-agents#developer-tools#paelladoc5 min./read →2026-07-11productI asked the Codex app to build an app twice. Neither made it to tomorrow.The same family manager, three times: the Codex app with no verification, the Codex app asked to verify, and PaellaDoc. All three repos are public. Only one makes it to tomorrow.#paelladoc#codex#ai-coding#vibe-coding#execution-gate12 min./read →2026-06-20productProduct knowledge graph: read your product's shape, not the repo's feature listClaude Code is excellent at reading a repo. To decide product, you need memory, provenance and the shape of what you built. That shape is what Compass reads.#paelladoc#product-management#knowledge-graph#compass#ai-coding10 min./read →2026-06-19productProduct management in the AI era: building got cheap, deciding got expensiveAI moved the bottleneck in product management. Building is fast now. Deciding what to build, with what evidence, under which assumptions, is the hard part. That is what Compass is for.#paelladoc#product-management#ai-coding#compass#evals9 min./read →2026-06-16productYou do not have a Claude Code problem. You are the runtime.Claude Code writes your code. Codex runs your tasks. The job nobody named is keeping the product contract alive until the work is validated, with evidence. Right now, that job is you.#paelladoc#ai-coding#agents#claude-code#runtime11 min./read →2026-06-14productNo vendor lock-in: route every task to the right engineMost AI products still ask which model to use. After enough broken tasks I stopped asking that. The question is which engine a task needs, and whether I can trust it to run.#paelladoc#model-routing#ai-coding#llm9 min./read →2026-06-12productThe product brain that maintains itselfI built a folder of context to feed my AI. It worked for two weeks. The hard part was never building it, it's who keeps it true when the code moves underneath you.#paelladoc#product-management#knowledge-graph#ai-coding8 min./read →2026-06-07productWhat is a feature factory?The term, the warning signs, why teams fall into it, and the one test that gets you out.#paelladoc#product-management#product-strategy#feature-factory4 min./read →2026-06-07productMaking product means installing a behavior, not running a feature factoryThe Bloomberg Terminal is ugly, looks like the 80s, and costs $32,000 a year. Nobody leaves it. The reason isn't the price, and it's everything about the difference between making product and shipping features.#paelladoc#product-management#product-strategy#feature-factory8 min./read →

— end of log · 9 entries in product —