Engineering and app development
JW Labs LLC builds software. Mostly we build it for ourselves — 4a and longlive are ours — and we take on outside work where the problem is the kind of problem we are good at.
If you have something to build, write to help@jwlabs.ai. That address reaches the people who would do the work, not a form.
How we deliver
No delivery layer between the person you email and the person who writes the code. No account managers, no relay, no handover. The work is heavily automated behind the scenes, which is what makes the path from a decision to working software short.
That shape is the offer. The conversation about what to build and the work of building it happen in the same place, a change of direction costs a conversation rather than a planning cycle, and you are talking to people who will still be here when the thing needs changing.
It also means we are not the right answer to every question, and we say which ones.
What we build
Four kinds of thing, described properly on the capability page:
Apps people use on a phone. Installable web apps that behave like native apps, and native iOS and Android builds from the same code as the web version — one thing to maintain rather than three.
Products with a lot of real data underneath them. Ingesting somebody else's feeds and APIs at scale, keeping the result correct, and turning it into something a person can navigate. This is most of what our own products are.
Software that has to be careful with people's data. Architectures where the privacy claim is a property of the design rather than a promise in a document — and the documents that go with them, including the ones an app store will not accept a submission without.
Front ends that are still working in five years. Sites and interfaces with no dependency graph to rot, no build server in the path, and nothing that breaks because a package six levels down was republished. This website is one.
The evidence
We would rather point at something you can open than describe ourselves.
4a is deployed and you can use it right now, in a browser, with no signup: open it. It is an installable web app that works offline, holds a large third-party catalogue, has no third-party code in it at all, and ships a working data-deletion control that reports honestly when it cannot finish.
longlive is live on its own domain — a scrubbable, filterable timeline over a decade of sourced material.
This website is the same taste at its simplest: no script, no dependencies, no external requests, built by one file.
What that adds up to is software you can open and inspect yourself rather than a description of it, and a page that says exactly what is built and what is not.
How working with us goes
The longer version is here. The short version:
You email the people who would build it. Same address as everything else on this site.
We tell you if it is not for us. Quickly, and with a reason. That happens more often than a company trying to fill a pipeline would admit to.
We agree what "done" means before we start. Not a document nobody reads — a short, specific statement of what the software will do, so that finishing is a fact rather than an opinion.
You talk to the people building it. Throughout.
We say what is not finished. Every time. It is the same standard the rest of this site is held to: we publish what is not built, we label what has been measured versus assumed, and we would rather deliver a shorter list of things that actually work.
When we are the wrong choice
Genuinely, and this is not modesty — sending you elsewhere costs us one email and sending you into the wrong engagement costs both of us months.
- You need staff augmentation — bodies inside your process, on your tooling, at your standup. That is a real service and it is not ours.
- You want a fixed-price fixed-scope build of something nobody has specified yet. We will not quote that, because the only way to hit it is to cut the part you cared about.
- The thing you need is design. We can build an interface and we care about how it feels, but we are engineers.
What we will not do in anyone's product
These are the same commitments our own products are held to, and they do not come off for a fee.
No dark patterns. No engagement mechanics designed to make leaving hard — no manufactured streaks, no infinite scroll where a list would do, no notification bait.
No advertising surveillance we would not accept ourselves. We will not bolt a tracking SDK into a product and then help write a privacy policy that talks around it.
No claim we cannot support. Not in the software, not in the store listing, not in the privacy policy. If a document has to say something is true, it has to be true, and if it is not yet, the document records the silence rather than filling it in. That is a real constraint on what we will put our name to and it is worth knowing before you hire us.
Get in touch
help@jwlabs.ai — one address for everything. Tell us what the software has to do and who it is for. A paragraph is plenty for a first email.