Philosophy
At Slytrunk we rebuilt how we work for what comes next in AI.
We are not running the same process with a model taped on. Roles, briefs, and how we ship have been reworked end to end. The point is not a little more speed on yesterday's backlog, it's what can actually be accomplished when the way you build matches what the tools can actually do.
Philosophy · The shift
This is not a small upgrade to the same machine.
For twenty years the whole machine was built around scarce developer time. Two-week slices, story points, ceremonies sized to how much a team could ship by hand. That made sense when building was the scarce part.
AI changes those limits. A draft that used to take days can land in an afternoon. What gets scarce instead is clear thinking — a written description of what to build, and judgment about whether the result is good.
The upside is not a faster trip through the same backlog. It is the work that suddenly becomes worth attempting.
Philosophy · The redesign
We rethought the whole development process — we didn't just add AI as another tool.
We changed how a job moves through the shop. Briefs carry more detail now. Design is what the agents build from. Nothing ships until someone here has checked it. The models sit inside that setup instead of off to the side while the week runs the same as before.
A new model bolted onto last year's workflow still produces last year's shape of work. We wanted the shop itself to change so bigger projects stop dying in the planning meeting.
Philosophy · Ambition
The work that used to die because it would take a year.
Plenty of good ideas used to die in the planning meeting. The build would burn the quarter, or the idea was too fuzzy to put in a contract. That math changed. Now you can put a working cut in someone's hands early and decide from the real thing.
That goes for client work and for our own products. The question we care about is simple: what would we have talked ourselves out of before that we can actually try now?
Philosophy · Where the craft went
The lasting work is the vision and the system that builds it.
When agents write a lot of the code, polishing generated lines like sacred craft is the wrong instinct. What lasts is the vision, and the harness that carries it: repo rules, tests, design as the source of truth, and house standards the model has to see every time, hanging together so the next pass inherits the context instead of drifting.
When the output is wrong, we go back to the system that produced it, fix what failed there, and run again. If you keep hand-editing the generated output instead, you end up maintaining something nobody on the team fully understands. We still write code by hand when judgment belongs in the lines, but that is not the default volume anymore.
Practices · Day to day
Agents need an environment, not a clever sentence.
A thin ticket produces thin output. Agents need the full picture: the conversation behind the ask, the design they have to match, the tests that fail if they get it wrong, and the house rules they have to follow.
People keep the hard calls: product taste, architecture, anything that touches money or trust. Someone here still owns the merge. Models wander. Rails exist because of that, not because we pretend the model has taste.
Practices · Quality
Nothing ships until someone here has had their hands on it.
That has not changed. We still put real constraints on the project and on each ticket, write tests that actually catch failures, and run another pass that can kill a change before it merges. "Impossibly Well Built" is still the bar.
We are not trying to remove humans from the loop. We are trying to stop spending human time on the volume that agents can do, so the people we have stay on the decisions that actually need them.
Contact
Got a problem with a real constraint in it?
Bring that. Or bring a person who should be in the room.