Founder @StateChangeAI. Youtube: https://nitter.cf/t.co/6PBjwzkS4q Statechange: https://nitter.cf/t.co/pxiIlMM5TD.
East Burke, VT
Joined July 2012
- Tweets6.7K
- Following1.2K
- Followers899
- Likes4.4K
Ray Deck @ statechange.ai retweeted
I’m have realized that more than half of my “let me just do this manually” actions had some kind of consequence that I didn’t foresee during editing. Edit the padding here, but forget it over there. Or now it breaks the responsive layout. The moment I delegated even these seemingly innocuous changes, things ended up with much fewer corrections required.
The biggest leap in developer leverage is moving from steering AI agents in real time to conducting them asynchronously through hierarchical rule systems.
pro.statechange.ai/c/weekly-…
Writing AI instructions as numbered legal clauses is the best way to fight context rot and maintain architectural control over complex projects.
pro.statechange.ai/c/weekly-…
“The big UI opportunity in AI is bringing your experience inside the flow.” Bring a smaller product surface into ChatGPT or Claude, where the customer and agent work.
Full essay: youtube.com/watch?v=5NRQWPrS…
“They don’t want another destination.” Customers want their agents to use your software in their own environment.
Full essay: youtube.com/watch?v=5NRQWPrS…
“You want to provide a smaller surface area.” A little visual, then let the customer and their agent decide how to work with your system.
Full essay: youtube.com/watch?v=5NRQWPrS…
🤖 Made with AI
ALT State Change infographic contrasting a screen-heavy software workflow with a smaller product surface used by a customer's agent, tool definitions, useful errors, and a compact visual result.
“The thread that I saw was the balance sheet.” Ray and Toby follow Miro’s announced deal from valuation to idle cash and investor pressure. Full conversation: youtu.be/SuZLAoRr2-E
Ray Deck @ statechange.ai retweeted
every product i use should just be a headless mcp, that feeds into an agent, where i never have to think about anything underneath.
except for our product, of course.
in our product, customers love the ui and love us very much.
we're special that way.
“VCs want you either to go crazy or die.” Toby on investor incentives when growth no longer supports the old multiples. Full conversation: youtu.be/SuZLAoRr2-E
“What do you think your valuation is?” Ray asks what Miro’s announced deal and Airtable’s sale mean for a smaller SaaS business still losing money. Full conversation: youtu.be/SuZLAoRr2-E
Share the idea that's still hazy in your head. Turning it into words makes your understanding firmer. Distill it to the core, and it may become easier for someone else to consume—so they can work on new ideas.
AI may restructure a market before it restructures the company inside it — AI-native entrants start from a lower interface and move value from vendor to customer.
Full conversation: youtu.be/MZwIHkmcDmg
Ray Deck @ statechange.ai retweeted
If AI had been writing and reviewing its own code for the past few decades, and humans had only recently started looking into what it was up to and discovered an ever-growing list of CVEs, we would reasonably conclude that AI hadn’t been good at coding all along and that we should just start over.
Except it’s the other way around.
Humans have been the ones writing and reviewing each other’s code for decades. Very diligently, if I may add.
If we’re being intellectually honest and holding ourselves to the same bar we hold AI to, we’d have to admit that we haven’t been all that good at it either.
Someone who works with a lot of Fortune 500 CEOs says their most common question about AI isn't strategic — it's "when do I get to retire?"
Full conversation: youtu.be/MZwIHkmcDmg
CEOs aren't just cutting costs — they're handing headcount to AI. The honest answer to "when do I get to retire?" turns out to be: reduce headcount, and let AI carry the load.
Full conversation: youtu.be/MZwIHkmcDmg
Frontier AI models are becoming interchangeable gears in a larger machine — the practical move is choosing the right model-plus-harness fit per job, not loyalty to one model, since new frontier leaders keep displacing the last favorite.
Full conversation: youtu.be/p6PijF6Vci8
“The token efficiency is much better — token cost seems to be lower because it's more efficient.”
Full conversation: youtu.be/p6PijF6Vci8
“It's actually the thing that surprised me the most about it is how fast it is.”
Full conversation: youtu.be/p6PijF6Vci8
Where does value survive when AI models get commoditized?
Specialist data, regulated context, or proprietary process — otherwise you become an invisible extension of someone else's prompt.
Full conversation: youtu.be/esvm0CLeT0s
