I just signed The No AI Slop Manifesto by @dexhorthy.
I’m refocusing my coding systems from ideas + visual design to reliable production code, and I’m looking for a new team to transcend slop and build epic things with.
Here’s what "no slop" actually means in practice 🧵👇
Still reliving this summer’s awesome generative UI and agent-to-agent collab with @petewsmart and the moment our late nights designing + building this vision paid off when he fist-pumped the air after the successful live demo @MIT@EmTechMIT Nailed it!
2 years ago at @MIT's EmTech AI, @petewsmart introduced generative experiences. This spring, he went back to show what comes next...
Agent-to-agent experiences.
In the full, live demo, a brand agent and a personal agent negotiate a unique interface on a shared canvas, in real time, for one person. This is the new internet.
Watch the full demo: fantasy.co/latest/the-new-in…
Over the coming weeks, we'll unpack 3 ideas behind it:
▪️ The interaction model
▪️ The privacy architecture
▪️ What happens when agents start acting alone
Have specific questions you'd like to see explored?
Let us know ↓
I helped a team evolve all the way from early inline autocomplete as a one-person shop to Claude Code and multi-agent swarms executing on defined system behaviors.
I’ve made mistakes, on runaway GitHub Action minutes spent re-testing untouched files to protect working behavior.
Real AI-native engineering means grappling with proportionality across human effort, energy, time, and cost.
Two things I’m actively refining:
• Token and action budget checkpoints (pausing to ask if and how an agent should keep going)
• CI guardrails smart to what’s changed
I missed the event where this was written, but I was there in 2025 for @swyx’s first AI Engineer Code talk on slop.
My takeaway hasn’t changed: the AI systems were always as strong as our human inputs. We scale ourselves and our teams by improving the system and its guardrails.