DHU Labs is a one-person software lab. You work directly with the person who designs and builds your solution, and serious AI collaboration does the heavy lifting, which is why the work stays affordable without being thin. You get real software running on your real data in days, not quarters.
Tell me what's slow, manual, or confusing in your business. I look at how things actually run and tell you honestly whether software would help, and if it would, I propose the smallest thing worth building. Then I build it while you watch it take shape, and I stay with it until it works in your real day rather than just in a demo.
Sometimes the honest answer is to build, sometimes it's to change course, and sometimes it's to stop, and you will always get mine plainly. I take on a small number of engagements at a time so that every project gets my full attention.
I build and run my own local inference hardware, so a system can be designed to keep your information on your own machines instead of sending it to an outside service. If you handle patient records, client files, or anything else that legally or practically cannot leave the premises, that is a design decision rather than a reason to stop.
Give it a pitch deck or a business description and it scores how replaceable that business is by AI agents, with a verdict: invest, investigate, or rethink. I built it because "I think AI could just do this" kept being my honest reaction in investor meetings, and a feeling isn't useful where a measurement is possible.
Live, currently invite-only. Ask me for access.Twenty questions that score a company's AI maturity across ten dimensions and return a summary you could hand to a board.
Live, currently invite-only.A chat assistant that reasons from a set of first principles instead of just generating plausible answers. In side-by-side testing it consistently beats the same underlying model running plain.
Early access.Recent projects include a revenue-intelligence tool that turns a sales team's emailed updates into margin-tagged deal records, a field-coverage app for enterprise field officers, and an automated pipeline that finds where four hundred production screens have drifted from their design files.
Named case studies available on request.I'm Joel Pennell. I started shipping products in 1999, in the place where music, computers and people meet: one of the first standalone consumer MP3 players, a home-stereo component that played CDs you burned yourself, back when most people still needed MP3 explained to them. Then storage systems for major music and film studios at Glyph Technologies, thin-client systems at Neoware through the dot-com years, early digital vinyl testing for Stanton, and a networked personal monitor mixer that debuted at AES in 2005.
After that I was the full-time parent at home for nineteen years, and I kept building the whole time. When AI collapsed the distance between thinking of something and building it, I came back to it properly, and I've shipped more than forty working projects in under two years. I work with several AI agents the way a builder works with power tools, and I've been shipping long enough to know the difference between something that demos well and something that holds up.
The name is an homage. DHU comes from the advisor who pulled me back into this work and first told me about the AI I now build with. Deeply Human Understanding is what the lab aims at: software that fits how people actually work, not how a system diagram wishes they would.
Describe what's slow or manual or confusing, and I'll tell you honestly whether it's worth building something, including when the answer is no.
hello@dhulabs.com