bellaifellow / hac — the practice path

Homebrew AI Club (HAC)

Build judgment through demos.

A working session for people to build their AI experience, expertise and judgment. We do it by doing the work and showing it to people who will celebrate your wins and tell you what's wrong with it.

Apply →

what is HAC?

Members demo their real setups, live: the agent, the pipeline, the prompt stack, the experiment they ran. We discuss what worked and what broke since last time, and we dig in together.

Judgment comes from running things, breaking them, and working out the kinks, with people doing the same. Why does one setup hold and another fall apart? The focus is what is on your machine and discussion, not lectures or slideware.

How it works: HAC trains judgment through doing, and the method is peer-to-peer. There's no instructor and no curriculum. You build something, show it to people building their own things, and the cohort's collective experience gets pointed at your problem — and yours at theirs. What you leave with is a sharper sense of what works, learned from people who tried it. We curate the participants to be peers who can teach as well as learn.

It's the Homebrew Computer Club, updated: a cohort of people with different skills, tinkering together, comparing notes on what works, and building discernment and taste for what deserves to grow versus what just sounds good.

two cohorts

The same format, calibrated to the cohort. The cohorts alternate weeks, so yours meets every other week — twice a month — and you're always with peers. Participants come from all levels from CTOs and CEOs of technology companies to individual consultants and freelancers, to students and recent graduates.

int

Intermediate

You use AI daily. Maybe you've done more than prompting like made a custom GPT, a script, or an automation. Demos of better prompts and tools are pitched so you can follow them and learn from them.

adv

Advanced

You run agents, pipelines, or loops every day. Maybe you're doing AI-based experiments and research. Demos go to full depth: architecture, failure modes, the parts that are still in development.

what it isn't

Not a course. No syllabus, no sequence, nothing to catch up on.
Not a lecture. Nobody stands at the front and explains AI to you.
Not a spectator sport. Everyone shows work eventually. That's how the cohort builds its taste — including yours.

a typical session

Three demos

~15 min each, live, the actual instance running (or breaking!)

What worked. What broke.

Everyone names one thing that paid off since last time and one that didn't. Both are useful: you find out what's worth copying, and what would have cost you weeks to learn the hard way.

Open discussion

Go deep on a specific project, get unstuck, or take on whatever just landed in AI — a new model, a new tool, a story worth arguing about

sections

Thursday evenings are the main run — the two cohorts trade off week by week. A midday video series is on the waitlist for people who can't do evenings.

Advanced — Thursday evenings Applications open

6:30–8:00pm · in person, San Francisco

Every other Thursday. Ten to twenty seats. Address comes with your confirmation.

Oct 1 · Oct 15 · Oct 29 · Nov 12 · Dec 3 · Dec 17

Intermediate — Thursday evenings Applications open

6:30–8:00pm · over video

Every other Thursday, on the weeks Advanced isn't meeting. Ten to twenty seats. Demos are screen-shared, which works well for this cohort.

Oct 8 · Oct 22 · Nov 5 · Nov 19 · Dec 10

Advanced — Wednesday lunch Waitlist

12:00–1:00pm Pacific · over video

Every other Wednesday, one hour: two demos instead of three. Opens if six people join the waitlist. No payment until it's confirmed.

Sep 30 · Oct 14 · Oct 28 · Nov 11 · Dec 2 · Dec 16

Intermediate — Wednesday lunch Waitlist

12:00–1:00pm Pacific · over video

Every other Wednesday, alternating with the Advanced lunch section. Same terms — it opens only if six people want it.

Oct 7 · Oct 21 · Nov 4 · Nov 18 · Dec 9

the basics

Cadence: Every other week — two sessions a month
Length: 90 minutes, hard stop
Format: Advanced in person in San Francisco; Intermediate over video
Cost: $180 for the fall series
Seats: Ten to twenty per cohort
Who's running it: Sunil Paul & Tenni Theurer
Part of: Bell AI Fellowship

apply

Each application is read, and admission isn't automatic. We're matching people to the right cohort and building a cohort where the demos are worth your evening, so places go by fit.

No interview, five minutes:

How you're using AI today
What you want to be able to use it for
A bit about your background
Your LinkedIn, if you'd like to share it — optional
Whether you'd demo, and which section fits

If the fee would keep you out, please say so in your application. We set places aside for exactly that, and it won't count against you.

Apply →

Opens the application form in a new tab — about five minutes.