Hi folks! I have a couple of good news :D Everything free I make now lives in one place: ainativesoftware.engineering. No account, no paywall, nothing to sign up for. Every page has a plain Markdown twin at the same URL with .md on the end, plus an llms.txt index. You can hand a page straight to your agent instead of copy-pasting from a browser.
Book news while I am here. All 11 chapters are written, about 135,000 words, and they are landing on the O’Reilly Early Release in the next week or so. Quite soon you will be able to read the whole thing instead of the first few chapters. Stay tuned, I will send a note the moment it is all up. Print edition is February 2027.
There are a couple of resources I already had and I’ve just moved to the website. The AI-Native Engineering Canvas is the workshop template: eight areas your team fills in together to design its own operating model. Print it or drop it into Miro, 60 to 90 minutes the first time, and it works async. If you never got round to running it, this is a good week to test it with your team. The 7-day roadmap is for someone starting from zero: 7 days, 40 topics, theory and a practice exercise for each, with the videos and articles worth their time.
Now the three new ones.
1. How agent-friendly is your repo? The Readiness Analyzer
A skill you drop into your repo with one command:
npx skills add alfonsograziano/skills --skill aine-readiness-analyzerThen you say: Run the AINE readiness analyzer on this repo. One agent, under ten minutes, 46 checks across five areas: context, specs, verification, tooling and safety. Safety is the biggest at 12 checks, which tells you where most repos bleed. It leaves an AINE-REPORT.md in your repo, so you get an artifact to re-run against later.
I ran it on this website’s own repo first and then on repos like Codex, Django LangChain and Supabase. From time to time it spots interesting things also on very well-known repos like these!
Every score and the full report behind it: ainativesoftware.engineering/readiness.
2. How well is your team actually working with agents? The Assessment
The analyzer only reads files. It cannot tell you that two of your engineers have opinions about AI they have never said out loud. So: 29 questions about what your team actually does, five areas, about ten minutes, no email and no account. Nothing is sent to a server.
You get a score out of 100, a radar across the five areas, and a question-by-question report built for a retro rather than for a screenshot.
ainativesoftware.engineering/assessment
3. What do you do tomorrow morning? The Baby Steps
Seven steps that make a codebase safe for agents, in the order to do them.
Tests. A safety net first, covering the paths that would hurt, run by one command.
Context. An AGENTS.md or CLAUDE.md, the rules, a few skills, so the agent stops relearning your repo every session.
Gates. Linting, types, secret scanning, dependency checks, and a branch that rejects anything failing them.
Specs. Intent written down before code, with acceptance criteria a machine can check.
Tools. The gh CLI, read access to deploys and logs, MCP servers for the tracker and the browser.
Orchestrate. More than one agent, in parallel or as a pipeline.
Scale. Take it to the team, then the org. Shared context, shared gates, an owner who keeps them current.
ainativesoftware.engineering/baby-steps
Where to go next
Audit your repo tonight, it takes ten minutes. Score your team at the next retro, together, arguing about the answers. Then find the lowest baby step you have not really finished and start from there :D
Audit your repo. ainativesoftware.engineering/readiness
Score your team. ainativesoftware.engineering/assessment
Pick your next step. ainativesoftware.engineering/baby-steps
Run the Canvas with your team, if you never did. ainativesoftware.engineering/canvas
Send someone the roadmap. ainativesoftware.engineering/roadmap
I like it so much those smalls.
Thanks
Doing great work, I will soon be launching my own book at 10xdevrel.atharvashah.com