Applied risk workshop
Westside Lab puts capital, code, and the lab's own time behind the odd, opinionated software the polished houses pass on. Half the bench gets killed. The half that survives earns a name, a brand, and the right to take up calendar space.
Working theses
Every specimen on the bench is downstream of these. We don't pretend to be neutral. Opinions are the point — and the polished houses are welcome to disagree.
A focused product that does one strange thing well will out-ship a platform that does ten predictable things. The cost of trying weird ideas has collapsed. The cost of generic ideas hasn't. We bet on the weird ones.
Examples on bench: mood-shifting digital art, signal in trash-tier data, software for the one-person business.
Every specimen starts with a serial number and a one-paragraph thesis. Names — and the press kits that come with them — are earned, not assigned. Half the work of a venture studio is resisting the urge to name a thing before it's a thing.
Average time from bench to name: 14 weeks.
We publish what we've killed alongside what we've shipped, with the cause of death attached. A studio that doesn't kill anything isn't running experiments — it's defending a roadmap. We'd rather be honest than tidy.
See the ledger.
Portfolio
A snapshot of specimens currently in active build or sitting on the bench. Nothing has a public name yet — codename, sector, status. That's the deal until something earns the right to launch.
Generative art · ambient
Generative digital art for the home that shifts with the room — mood, time of day, weather. A speculative bet on whether decor should be static or responsive. Currently rendering on three living-room screens.
Pipelines · weird signal
Most useful information is the kind nobody bothered to clean. We're running pipelines that find signal in receipts, group chats, dead forums, and the stamps on the back of photographs. Buyers TBD.
Small-business software
Software that runs an entire small business with one human in the loop. Inventory, billing, replies, taxes, and the polite robot that nudges the accountant on Friday. On the bench while we recut the data layer.
UX · AI
Interfaces that build themselves around the question you're actually asking. Not chat. Not a dashboard. A UI that materializes for the job and disappears again. One client is shipping it behind a flag.
Personal software · weird
Treating mood, ambient sound, and time of day as legitimate inputs. The hypothesis: personal software gets better when it knows how the room feels. Early, deliberately. We don't know if there's a customer yet — that's the bet.
How the lab works
Westside Lab is not a fund. Not a holding company. Not a quarterly partners deck. It's a bench, a list of hypotheses, and a small group of people who like to ship software that other people thought was too weird to back.
Every specimen starts as a one-paragraph hypothesis with a sector, a target user, and the failure mode we're betting against. If it can't survive ten minutes of argument, it doesn't get a slot on the bench.
Two to six weeks, one operator, one engineer, no roadmap. The prototype is the spec. If a hypothesis can't be built that fast, it was probably too broad — and almost never gets better with more time.
Codename only. We ship to a small list, watch how people actually use it, and listen for the surprise — the move we didn't design for. That move is usually the product.
Most specimens get killed. A few sit on hold. The rest graduate: they get a name, a brand, a domain, and the right to take up calendar space. The kill decision is published the same week it happens.
Every specimen leaves a writeup — cause of death or notes on what shipped. We're a lab, not a graveyard. Killed ideas teach the next one, and the next one almost always works better. The notes are public.
The kill log
Most studios publish what shipped. We publish what didn't — with the cause of death attached. If a venture studio has a kill rate of zero, it's either lying or it's not running experiments. We'd rather be honest than tidy.
Lab notes
A short log every couple of weeks — what shipped, what got killed, what's about to escape the bench. No newsletter sales pitch. No drip sequence. Just the field notes.
One of the two active builds. New compositor on the back end means a fresh frame renders in under half a second on the living-room screens. The bet is starting to look less stupid. Wider pilot next month.
Eight weeks in. The data was there, the demo was there, the soul was not. Notes are filed. The lessons paid for WSL · 004. (Full writeup in the ledger.)
Watching someone use a UI that we didn't draw was the best afternoon at the bench in a long time. One client, one flag, ten users. Going wider in Wave 10.
Treating mood, ambient sound, and time of day as legitimate inputs to personal software. We don't know who buys this yet. That's the whole bet. Moving WSL · 005 from bench to active when the prototype lands.
// no tracking · no drip · short log every two weeks-ish //
Direct lines
Tell us what you're building. We answer fastest to interesting problems, useful introductions, and people who've already shipped something we liked. We're not selective in the precious sense — we're just small.