Proficient Android AI
A program to make Android engineers more proficient by giving them the tools and knowledge to guide their AI agents the right way.
50 seats left
$399 per attendee · secure Stripe checkout · no account needed
Jorge Castillo
Android engineer at Disney · 15+ years
15 years of Android engineering, including at X and Disney, on the codebases where architecture actually matters.
Most teams rely on AI agents to generate code daily, but not many of them guide their agents correctly. I have done an intensive work of separating what genuinely makes an agent productive for Android codebases from what merely burns tokens or generates slop, and packaged it into a set of hands-on topics.
Same team, same agents, half the waste.
Makes them significantly more proficient as Android developers by teaching them how to optimize and direct their AI skills for more on-point results.
up to 50%
fewer tokens burned
Same work, half the spend, on any state of the art coding agent.
1 codebase
one set of conventions
Agent output that matches your architecture and guidelines instead of fighting them.
0 slop
PRs you can defend
Every change understood, curated and auditable before it reaches a reviewer.
11 modules, all hands-on
Every module ends with something your team keeps: a configuration, a skill, an audit, a prompt set that works against your own codebase.
- 01Tokens
Cut token spend by up to 50%
How to save up to 50% of the tokens when working in Android projects with any of the state of the art AI agents.
- 02Context
Agents that respect your architecture
How to guide and optimize the output of your coding agents to match your Android codebase architecture and guidelines.
- 03Android CLI
The Android CLI and the official knowledge base
How to leverage the Android CLI for streamlined access to convenient Android Skills and the official Android knowledge base, including the official docs.
- 04Delegation
Delegate the heavy lifting
How to delegate to AI agents the most common tasks in Android development: building UI components, UI migrations from Views to Compose that follow all the Compose guidelines, migrating navigation libraries, moving from fragments to full Compose, and more.
- 05Proficiency
Better engineers, not dependent ones
How to effectively improve your proficiency as an Android developer thanks to AI, instead of quietly outsourcing your judgement to it.
- 06Determinism
Deterministic results from LLMs
How to achieve deterministic results when coding Android apps with LLMs, so the same intent produces the same quality of output every run.
- 07Audit
Audit everything that lands in your repo
How to audit any changes committed to your repo following the official Android guidelines, and catch any violations of a given set of requirements.
- 08Skills
Android Skills that pay for themselves
Android skills that can really save your team time and avoid future regressions, packaged so the whole team benefits from them.
- 09Tests
Tests that survive a refactor
Most generated tests are unit tests covering a single class: tied to implementation details and prone to fail with every minor refactor. Learn to generate tests that cover realistic flows from a user perspective and resist API changes and refactors.
- 10Team flow
No more slop PRs
How to use AI agents in a productive and efficient way for your team, to avoid massive amounts of slop PRs that nobody, including their author, actually understands.
- 11Review
Understand before you commit
How to understand and curate all the changes produced by your AI agents before committing them or sending them for review. Being respectful to your teammates.
Built for teams already paying for agents
- Android teams already paying for coding agents and unsure what they get back.
- Engineering leads watching token bills grow faster than shipped features.
- Teams mid-migration: Views to Compose, fragments out, navigation libraries swapped.
- Senior Android engineers who want AI to sharpen their craft, not replace their judgement.
50 seats, numbered as payments clear
The cohort stays small on purpose, so everyone gets time on their own codebase. Seats are numbered in the order payments come in.
50 seats left
$399per attendee
One seat in the next cohort, starting January 7, 2027. All 11 modules, worked through against real Android codebases.
No account to create. Pay through Stripe with the email you want the cohort details sent to. Your seat is confirmed the moment the payment clears.
- 11 hands-on modules
- Skills, audits and prompt sets you keep
- Agent-agnostic: the techniques travel with you
- Work against your own Android project
- Capped at 50 attendees
Before you register
Which AI agents does this cover?
The techniques are agent-agnostic and apply to any of the state of the art coding agents. The workflows, skills and audits you build during the program travel with you if your team switches tools.
Is this a talk or hands-on work?
Hands-on. You work against real Android codebases, wire up skills, run migrations and audits, and leave with the configuration and prompts your team will keep using.
Do I need to prepare anything?
An Android project you are allowed to work on, and an agent subscription your team already uses. Everything else is provided before the cohort starts.
How does the 50% token saving work?
It comes from context discipline: what you load, when you load it, how you scope tasks, and which work you delegate to sub-agents versus handle inline. Most teams overspend on context they never needed.
Can my company pay for several engineers?
Yes. Check out once per engineer, using each attendee's own email so the cohort details reach the right person. For a single invoice covering several seats, get in touch and it will be handled off-platform.
Do I need to create an account?
No. Pay through Stripe with the email you want the cohort details sent to, and your seat is confirmed as soon as the payment clears. There is nothing to sign up for and no password to remember.
Which email should I use?
Whichever one you actually read. Everything for the cohort (joining details, prep material, the schedule) goes to the email you pay with, so use that engineer's address rather than a shared billing inbox.
50 seats left for January 7, 2027.
A program to make your Android engineers considerably more proficient by leveraging the power of AI agents the right way.