Castleridge Labs
AI research and software studio.
We turn repeated work into focused products people can rely on — then carry what we learn into the next one.
Studio thesis
Research through products, not demos.
Our research starts with repeated work: the habits, decisions, and handoffs held together by memory, tabs, and workarounds. We study one loop, build the smallest complete product that closes it, and learn from how people actually use it.
- 01
Noise
The same job done a slightly different way each time.
- 02
Pattern
The shape of the loop becomes visible and repeatable.
- 03
System
A shipped product runs the loop the same way, every time.
- 04
Compound
What it teaches carries into the next product.
Products
Chapter 01Live
James Peralta
Make the hard part familiar.
Stated on jamesperalta.com
Interview preparation that compounds instead of resetting.
- Learn
- Teach
- Interview
- Apply
Interview prep normally starts from zero every session — new notes, new questions, no memory of what went wrong last time. The product turns that into one continuous loop: study a concept, explain it back, rehearse it under interview conditions, then carry the same context into the application.
The public site organises the entire product around four moves — Learn, Teach, Interview, Apply — and leads with a single promise: repetition is what makes difficulty ordinary.
Open James Peraltajamesperalta.com
The product in motion — reasoning, follow-ups, and code in one interview surface.
Chapter 02Live
UFC Calendar
Never miss a fight again.
Stated on ufccalendar.tech
The fight-night lookup, done properly once.
- Card
- Night
- Result
Checking when a card starts, which fights are on the main card, and what happened afterwards is a lookup people repeat every few weeks — usually across three unreliable sources. This is that lookup collapsed into one browser-native surface that sits where the fan already is.
Shipped as a browser extension rather than another destination site: the schedule appears in the tab you already have open, with the main card and prelims split the way fans actually read them.
Open UFC Calendarufccalendar.tech
The lookup, in-tab — local start time, matchups, card tabs, and spoiler control.
Chapter 03Live
AI VAs
One inbox, every enquiry answered.
Stated on aivas.cloud
AI Virtual Assistants for business chat that keep the conversation going.
- Enquire
- Reply
- Review
- Book
Business enquiries arrive in Instagram DMs at all hours — openings, prices, bookings — and go cold when nobody is free to answer. The product closes that loop in the inbox where it already happens: the assistant replies in the operator’s voice, escalates anything off-script for human sign-off, and carries a qualified lead through to a booked appointment.
The public surface leads with a single promise — one inbox, every enquiry answered — and shows the loop as four moves: answered DMs, a human review queue for off-script asks, logged conversion, and a calendar confirmation instead of another to-do.
Open AI VAsaivas.cloud
The loop in the inbox — reply, review, convert, and book without leaving the chat.
The connective tissue
Three products, one method.
One loop, closed end to end
A product earns its place by finishing a loop, not by covering a category. Learn → Teach → Interview → Apply. Card → Night → Result. Enquire → Reply → Review → Book.
Meet the loop where it happens
Interview prep lives on a study surface. A fight lookup lives in the browser tab. Business chat lives in the Instagram inbox. The shape of the product follows the shape of the habit.
Scope is a decision
Every feature we skip is something we chose not to get wrong. Narrow is the strategy, not a limitation.
Research earns its way into products
Models and methods change. The durable work is learning which workflows deserve automation, where judgment belongs, and what people will trust.
Different products. One standard: make the outcome feel inevitable.
Castleridge Labs researches recurring work by building focused software around it. The clearest proof is what we have shipped.