AI product company · est. 2026

Do it, machina.

Deus ex machina, minus the deus.The old stage trick where a machine swoops in and fixes everything. Here the machine just does the work, and something checks it.

doitmachina builds AI products for real jobs: computer programs that look at pictures and write words, aimed at everyday work. The machines do the work. The checks keep them honest.

Need playback or device help? David also takes contract work. Hire David →

Products

Three products in the works

Each one is working code at the stage its label says, and no further.

slideinto.house

Local beta

Listing films, without the film crew

A broker drops a listing folder into one box, confirms the facts, picks a voice and approves the script. Out comes a narrated film with captions, read from the broker’s own copy.

You do

Upload the listing folder (with a floor plan), confirm the facts, pick a voice, approve the script.

You get

A narrated film with captions, from your own copy.

Free during the beta: the first walkthrough is watermarked, and picking a plan unlocks the clean film, also free for now. No payments are taken yet.

  • Virtually Staged
  • Concept – Not Included
  • AI-enhanced mark
  • Claims checked
How it stays honest, and how it works
  • Staged rooms are badged Virtually Staged, concept ideas Concept – Not Included, and AI-generated shots carry an AI-enhanced mark. The end card repeats the disclosure.
  • A second check compares every claim in the script to your listing facts, and every on-screen label to its photo.
  • By default, motion is a gentle depth effect on your real photos, so it never shows a room that doesn’t exist.
  • AI sorts the photos and suggests the shots. Paid AI steps run only after the budget is approved.
  • Local beta means it runs on doitmachina’s own machine while it’s tested. It isn’t a self-serve website yet.
  • Check your local listing service’s (MLS) rules on AI-edited media before publishing.

Ask about the beta Tell us your MLS and the kind of listing. You’ll hear back from David directly.

Illustration: a listing film frame with a caption and a Virtually Staged badge VIRTUALLY STAGED Three bedrooms, one very good porch.
Illustration of a listing film frame

Spot-Canon & Slop-Canon

In development

Tap the video. The buddy talks back.

Press on something in a video, and an animated buddy tells you what it is, grounded in what’s actually on screen. Two products from one codebase, with four fox characters: Piper, Ember, Ripple and Moss.

Two separate products, not a setting. Spot-Canon: kids 3–12, not open to families yet. Slop-Canon: teens 13+, starting as a private invite alpha. By design they run as separate deployments with separate data, sessions and catalogs, so a child can’t drift from one to the other.

  • Taps on pre-recorded video only
  • No microphone, camera or typing
  • No child name, account or profile
  • Nothing remembered between sessions

How the buddy talks: only the tapped part of the video is sent to an AI service to identify it; the buddy’s short line is then generated and turned into a synthetic voice. Nothing about the child is sent. Any future sign-in gets its own privacy and consent review first.

Spot vs. Slop

Spot-Canon

  • For families, ages 3 to 12, with family-safe personalities and a small reviewed catalog (one title so far).
  • Not open to families yet. Before it is, it needs a parent notice and consent flow and a legal review under COPPA, the US law protecting children’s privacy online.

Slop-Canon

  • Teens 13 and up: snarkier buddies, mild language, and propeller beanies. A private invite alpha is planned. Never offered to families.
  • Still bounded by written policy: no sexual content, glamorized substances, slurs, real-world politics or jokes about real tragedies.
Piper, an orange cartoon fox with a teal neckerchief, smiling
Piper, one of four fox characters (early artwork)

White-Cold Stonks

Research

Stocks the news forgot

One question: which stocks are moving without much news, and what usually happens next? News attention and price behavior side by side, in plain language. Not a broker, a trading system, or investment advice.

Where it stands
  • Testing whether a burst of news attention is followed by a cooling period and a partial price correction.
  • The first proof of concept said “I don’t know”: the news archive had too many gaps, so it refused to draw a conclusion. That’s the method working.
  • Next: a careful test that runs forward in time, collecting on a fixed schedule, so nobody can fudge the results. Nothing gets published until real data could prove the idea wrong.
Illustration: news attention spikes and fades; a gap in coverage is marked not evaluable NOT EVALUABLE news attention price
Illustration: a coverage gap is marked, not guessed
How it’s built

The machine does the work. Something checks it.

Check what it writes

Generated text is checked before it ships: listing scripts against the confirmed facts, buddy dialogue against written policy. What can’t be supported is cut.

Label what’s generated

Staged, concept and AI-assisted material is marked where it appears, not buried in a footer.

Say “I don’t know”

When the data isn’t good enough to answer, the product says so instead of guessing.

Privacy as architecture

If a product doesn’t need a microphone, a camera or a profile, it doesn’t have one. Anything more gets its own review first.

Honest status labels

Beta means beta. Research means research. Nothing here is dressed up as shipped.

Spend on purpose

Paid AI steps run only after the spend plan is approved, and every job has a cost report.

Founder

David Crandall

Founder & Principal Engineer

24 years of software, about 20 of them getting video onto TVs, set-top boxes and streaming apps, most recently at Charter / Spectrum. Founded doitmachina in July 2026 to build the products above.

Also available for contract work alongside doitmachina: streaming playback, devices and DRM. Typical start: send a few affected sessions or logs, and get a written finding back.

David’s résumé
Contact

Talk to David.

Early access, partnerships, or a hard problem. One inbox for all of it.

Email doitmachina