Float Point works directly with businesses to find and fix what's holding them back — and builds AI products that run entirely on hardware you own.
Most businesses don't need another off-the-shelf product. They have a specific, awkward problem that no existing software was built for — and what they need is someone technical to understand it and build the thing that fixes it.
That's the core of what Float Point does: consultation and hands-on technical work, shaped by the problem in front of you rather than by whatever we happened to build already.
How it worksA conversation about what's actually slowing the business down — not a sales call.
We find where the friction really is, which isn't always where it looks like it is.
Then we build it, integrate it, and make sure it holds up in real use.
LetsBuySpiritAir was covered across the news and drew millions of views on social media. Jonathan Hatch took it over as Lead Engineer and was the only technical person on it, serving millions of requests over a single week. Efficiency isn't an abstract principle here — it's the only reason that was possible.
Read the storyNothing Inc. works on the hard efficiency problem: making large models small and fast enough to run outside a datacenter. Float Point turns that research into software you can actually use. Neither product is generally available yet — Ivan is being tailored for its first pilot, and Classic IDE is still in development.
A locally hosted AI receptionist and assistant for professional practices. Ivan answers calls, screens and intakes new clients, books appointments, and keeps records organized — without a single client file leaving your office.
An AI-assisted coding environment for macOS and Windows that feels like a datacenter is behind it — while running a multi-agent framework on the GPU already in your machine. Go fully offline whenever you want.
Privileged client information does not belong in someone else's cloud. What we build runs inside your own network, on hardware you own and can walk over to.
Nothing Inc.'s research makes models dramatically smaller and faster without giving up accuracy. That efficiency is what lets a local product be genuinely good, rather than the compromise you settle for.
Running offline isn't a stripped-back backup mode. If your connection drops — or company policy means you never had one — the software behaves exactly the same.
If something technical is holding your business back, that's the conversation we want to have. Tell us what's in the way.
Get in touch
