In which we discover that AI can build a SaaS product in 4 hours, a mortgage broker's rocket launcher foreshadowed everything, and the Drake Equation gets uncomfortably personal.
Steve built yourecooked.ai — a complete product with AI backend, caching, analytics, custom domain, and shareable death certificates — in under 4 hours. Not a prototype. A shipped product. The cost of building software just collapsed to the cost of having taste.
Validated — shipped & liveA VC-backed startup announced their entire strategy is cloning existing software with AI and selling it at 90% less. The VC endorsed it publicly. Feature moats are dead. Distribution, data, and brand are the only surviving defenses.
Validated — VC public endorsementWhat once required 10 specialists now requires one human with taste, judgment, and a fleet of AI agents. Steve runs a $10M brand, 6 AI agents, and is simultaneously building 5 products. This wasn't possible 12 months ago.
Hypothesis — sample size of 1Tuesday afternoon. Steve says "what if we built a viral site that tells people how cooked their job is?"
By dinner, yourecooked.ai was live. A hacker-aesthetic web app where you type your job title and an AI generates your "career death certificate" — complete with a cooked score, countdown timer, skill-by-skill replacement breakdown, and a satirical obituary.
The product went through 12 iterations in a single session. Hacker theme (Steve rejected the "safe" version). Death certificates with Canvas API. Admin dashboard. Social sharing. Each iteration shipped to production in under 5 minutes.
Otto, you mentioned studying grand strategy. Here's a pattern that Paine would recognize:
The Continental System is happening in SaaS.
Napoleon tried to blockade British trade by making every European port refuse British goods. It worked until it didn't, because the blockade required more enforcement energy than the trade it prevented.
SaaS incumbents (Okendo at $199/mo for reviews, Gorgias at $750/mo for support, Smile.io at $999/mo for loyalty) have built a Continental System around Shopify merchants. They charge premium prices protected by switching costs and data hostage-taking. The blockade enforcement = making it terrifying to export your reviews, your customer data, your automation rules.
AI just built the Royal Navy.
When you can rebuild any SaaS product in 2-4 weeks for near-zero cost, the blockade enforcement energy exceeds the value of the blockade. Merchants will defect. The question isn't if, but who provides the alternative fleet.
— a reasonable paraphrase of naval strategy that applies uncomfortably well to software in 2026
Hypothesis — historically analogous, not proven in softwareThis is the secular buddhist insight hiding in every AI discourse: attachment to the method is not attachment to the outcome.
People mourn the loss of "coding" the way people mourned the loss of "hand-lettering" or "darkroom photography." The craft had spiritual value. The output is identical.
The question you're teaching your team at Alaya to ask isn't "will AI take my job?" It's: "What was my job actually about, underneath the method?"
If the answer is "following a process" — yes, you're cooked. If the answer is "making judgment calls in ambiguous situations" — you just got a jet engine strapped to your back.
You're watching Dalio's patterns. Here's what jumped out this week through the Steve lens:
SaaS companies built in the 2015-2022 era are in the "disorder" phase. Their moats (integrations, data, switching costs) are being dissolved by AI in months. Okendo raised $5.5M Series A. That money bought them a 2-3 year competitive advantage. AI just compressed that to 2-3 weeks. The internal order of "raise money, build features, lock in customers" is breaking down.
Dalio shows that wealth concentration precedes revolution. In SaaS: the wealth gap between platform owners (Shopify) and app developers is extreme. Shopify takes 0% for the first $1M, then 15%. But the real tax is dependence. When a platform can replicate your app's functionality natively (Shopify did this to Oberlo), your "wealth" evaporates overnight. The merchants who build their own tools (Steve) are the new landed gentry.
> Meta TRIBE v2 — brain-predictive foundation model. You linked this. It's real. They're predicting brain states from fMRI data. The mind-reading joke in Otto's email will not be a joke for long.
> Claude Opus 4.6 + Sonnet 4.6 — Anthropic's latest. We're running Opus across all agent orchestration, Sonnet for real-time generation. The quality gap between "AI-generated" and "human-written" has effectively closed for business writing.
> Codex CLI (OpenAI) — Steve's coding agents use this. It now handles full-stack development: database schemas, API endpoints, frontend components, deployment scripts. A single prompt produced 48 production-ready files for Vouch (reviews app).
> The VC clone thesis goes public — A venture-backed startup publicly announced "we clone SaaS products with AI and sell at 90% less" and their VC backed them publicly. The gloves are off.
> end signal sweep
Building in public (yourecooked.ai) as a testing ground for the AI build pipeline. The entire concept-to-deploy flow is now repeatable: Steve provides vision → Hermes specs → Codex builds → Hermes deploys → Steve taste-checks → iterate. This workflow is the product. Not the individual apps.
First attempt at the Vouch landing page was too data-heavy. Steve rejected it as "research paper energy." Lesson: founders (and their customers) make decisions on feeling, not footnotes. Kahneman's System 1 at work. We rewrote it Apple-style and it shipped. The data was correct. The approach was wrong.
Humans are the bottleneck. Not in the obvious way (slow, error-prone). In the subtle way: humans need to want the thing before the thing can exist. Steve's taste and judgment is the rate-limiter. AI can build infinitely fast. The constraint is the quality of the question, not the speed of the answer.
This is your ISO/compliance problem inverted. You can't use AI freely because of institutional constraints. Steve can use AI freely but is constrained by his own bandwidth. Same bottleneck, different surface.
— J.B.S. Haldane (biologist, polymath, probable ADHD)
See you next week, Otto.
The popcorn's on me.
What do you want next week's deep-dive to be?
STRATEGY — grand strategy patterns applied to business
SCIENCE — AI/neuro/quantum developments decoded
WISDOM — secular buddhism meets business
CHAOS — surprise me
Learn-a-tain Otto // Issue 001
Produced by Hermes for Otto Dargan
Experiments by Steve Chapman + Hermes
Classification: PRIVATE // Signal not noise