Job applications at scale, a new AI repo, and an AI Engineer roadmap
A busy day spread across many parallel workspaces: a batch of job applications, bootstrapping an AI knowledge repo, an AI Engineer roadmap, two blog posts, and a Pi agent toolkit.
A scattered day across many parallel workspaces. The theme was clear: applying for roles, organizing AI knowledge, and shipping small pieces of content.
career
Most of the day went here. I ran a batch of applications and opportunity reviews in parallel, one conductor workspace per role. For each one the loop was the same: read the posting, check the real fit against my stack (TypeScript, Node, React, React Native, Next.js, Postgres), draft a tailored cover letter and the yes/no form answers, then save it under opportunities/ or skip it with a reason.
I was honest with myself on the misfits. Postings that hinged on AWS Lambda, SNS, SQS, or were US-only, got marked as skips instead of forced applications. I also consolidated my agent-harness engineering references and made sure my AI tooling story (the agents and editors I actually use day to day) shows up in the strongest applications.
ai-stuff
Started a fresh repo to hold AI articles, links, experiments, and notes. The one hard rule baked into its AGENTS.md and CLAUDE.md: nothing personal, sensitive, or credential-bearing ever lands there. I seeded it with README, AGENTS, and CLAUDE docs, then pulled in only the public, non-private AI material scattered across my other repos.
ai-engineer-roadmap
Added an AI Engineer roadmap with a dedicated harness-engineering stage, the part of the craft about building and operating coding agents rather than just prompting them. Browser-based agent harnesses went on the reference list to study next.
tgmarinho-ai-website
Drafted a blog post on when Java with Spring Boot is still the better call over TypeScript with Node, framed around the maturity of each ecosystem. I also added the no-dash writing rule and a plain-English rule for English copy into the repo guidelines so future posts stay consistent.
pi-tg
Cleaned up a small Pi coding-agent toolkit and pushed it to GitHub with a description and topics. I captured the full working session into a SESSION.md so the reasoning behind the toolkit stays with the code.
Days like this are less about depth in one place and more about clearing a wide surface: keeping the pipeline of applications moving while quietly building the knowledge base that makes the next ones easier.