If you're nodding at three of these, you're in the right room.
There's no answer to any of this yet — not from a vendor, not from a framework, not from the field.
Anthropic's co-founder and CEO, Dario Amodei, says AI will eliminate coding first. Kent Beck — who's shaped how software gets built for decades — pushed back hard: that take "comes from someone who does not understand what software engineering actually is." His own verdict on where we are? "Nobody knows. It's not just that I don't know — nobody knows."
Kent Beck on The Pragmatic Engineer, 2026 — pushing back on Anthropic's Dario Amodei
When the people building the models and the people who wrote the methodology can't agree on what AI does to delivery, that's the tell: there is no playbook yet.
~15 yrs
roughly how long a disruptive technology takes to settle into its working patterns. We're at the very start of that clock.
Wait fifteen years for the patterns to settle — or start writing them now.
You've likely got as many years in the trenches as I do — building, optimizing, and bulletproofing delivery. The difference isn't the experience. It's that pointing all of mine at this one question is now my whole job.
At First Connect I lead the AI Platform team with a single mission: figure out how AI actually changes our SDLC, and help my director prove to the board that it works. I run that experiment every day — the part you can't, because you've got an org to run.
But one org isn't enough to write a playbook the whole field can trust; that takes a handful of sharp directors testing it in parallel. So I'm assembling a fellowship — a small room figuring it out together, in real time.
— Maks · ex-Head of Engineering, now AI Platform Team Lead at First Connect · 4Developers speaker
Better pull requests. Stricter gates. LLMs reviewing LLMs. More AI-written tests you can't fully trust. Bolting ever more checking onto a process AI already outruns.
GitFlow, PRs and code review are the wrong foundation for AI-speed delivery — you can't out-review an agent. Make the pipeline itself safe to move fast instead. It's what the giants proved decades ago, and the only base I see for an autonomous AI software factory.
Four practices, working as one system — they make speed safe without a human gate in the loop:
Executable specs, tiny batches, verification and delivery on autopilot — a pipeline an agent can run at full speed and stay safe. That's the factory. A strong opinion, tested daily; if it's the bet you'd make too, you'll fit this room.
Four directors to a room, one hour a week for twelve weeks — and I'm in every room. Small enough that everyone actually gets airtime. We don't trade opinions in the abstract — we read real pipelines and leave with real moves. To make that work I built Cadence: a local-first tool that reads your delivery pipeline — where the time goes, how much is AI-assisted — so you walk into the room already holding a dashboard of your own delivery — not theory. What I learn and the tools I build myself come to the room; you bring your org's reality.
Hiring someone to crack this full-time for your org costs a serious salary. A founding seat costs $49 — and you're sitting beside someone already doing the work every day.
$49 holds one of eight founding seats. You get my calendar on the spot to book your setup call.
One hour with me, one-to-one. We point Cadence at your repo and you walk away with a live dashboard mapping your SDLC to the metrics that actually matter — your numbers, on screen, not theory.
One hour a week with three other directors — four to a table, so nobody just lurks. We read real numbers; you leave with a move.
The dashboard you build with me in your first session — your SDLC mapped to the metrics that matter. The short course, for the theory behind it, comes with your seat to dip into whenever you want.
The room, the access, and the instrument to feed it. $49 holds your seat — founding price, first eight only.
You're not buying two products for $49 — you're reserving a seat in the first rooms, at a price only the founding eight get. The tools come with it; the room is the point.
That's this cohort. The next rung is Blueprint — the cockpit you hand your engineers for spec-driven work with Claude Code: a shared backlog that briefs the agents, parallel sessions you watch and steer, every change attributed back. Founding members get first access. Not part of this step — just where the road goes.
A short tour of the cockpit — backlog to briefed agents to attributed diffs.
My team could build this dashboard in a weekend.
Probably — the charts are the easy part. The hard part is knowing which metrics matter for AI-assisted delivery, how they move together, and what a given reading tells you to do. That's the room and the course, not a weekend of D3.
Is this just another talking shop?
No. Every session puts a real pipeline on the screen — someone's actual numbers — and you leave with a specific move. Not venting, not vibes.
I don't have a spare hour a week.
It's one hour, and it replaces the hours you lose guessing at this alone. Miss one — office hours and the recording. It buys time back, it doesn't cost it.
Only $49 — and you're still figuring it out too. What's the catch?
No catch. Nobody can sell a finished playbook yet, so I won't price like it's one. Hiring someone to crack this full-time for your org costs a serious salary; a founding seat costs $49, beside three sharp peers and someone already doing the work daily. It goes up once the founding rooms fill.
Two rooms of four — small enough that everyone gets heard, and I run both myself. If more than eight of you show up, one more room; then that's the cap.
This low only for the founding rooms. Once they fill, price and format both change.
The deeper program, the ongoing room, and founding Blueprint access come after — and cost more. This is the way in.
Reserve for $49, book your onboarding call on the spot. If the cohort doesn't fill or the first session isn't for you — full refund, and you keep the call and Cadence. You can't be out anything.
Two founding rooms of four engineering directors — one live hour a week each, plus office hours, led by me. Small tables, so everyone gets airtime. We work your real pipelines, not slides. Cadence and the short course come with your seat.
You get my calendar on the spot and book your onboarding call for whenever suits you. On that call I set Cadence up on your repo, so you arrive at session one with your own data on the screen. Your room kicks off once the founding seats are filled.
One hour a week for twelve weeks, plus optional office hours. Your run starts once your room fills. Miss a session and there's a recording. It's built to give time back, not take it.
You get a full refund — and because I onboard you before the room starts, you keep the call and your Cadence setup regardless. Reserving early is pure upside: worst case you're out nothing and you've still had a personal setup session and got the tool. Same deal if the rooms fill but the first session isn't for you.
Proprietary and source-available — a bonus with your seat, and I onboard it with you. It aggregates to the team or cohort and never surfaces per-engineer numbers, following the SPACE authors' guidance that individual-level throughput creates the wrong dynamics.
It's the founding price for the founding rooms. You're helping write the playbook and I'm learning from your org in return — the price reflects that trade, and it goes up once they fill. Far under any approval threshold; receipt available.
The deeper program, the ongoing room, and founding access to Blueprint open up — at a higher price. Founding members get first call. This seat is the way in.