I’ve been saying it for a few months: expect gaming to freeze from a hardware standpoint. Consumers are not going to be able to purchase upgrades. New consoles will cost $1000+. Gaming GPUs even more. The trend is going towards gaming PC, or reusing your work PC or Mac.

Just submitted Chiptunes to the @OpenAIDevs WebMCP hackathon. It's a Game Boy music app your agent can drive. It's based on a real Game Boy emulator in your browser and and it composes Game Boy accurate music instantly and and offline! You can tell the agent what kind of music to do and and what kind of moods and and you can even export your music as Wav or Game Boy ROM! What an agent can do with it: 1/ "Write me a dungeon theme like Castlevania, 40 seconds, no drums" — and it plays, in the tab you're both looking at. 2/ Ask for twelve complete, different songs. They come back in about 70 ms. Breadth is free when a song costs 1.6 ms. 3/ An agent can't listen, so it measures instead: mode, phrase shape, consonance, density. It checks its own work. 4/ Export the result as MIDI, a share link, or a 32 KB .gb cartridge that boots on a real Game Boy. Everything is deterministic and open source (MIT). The music comes from a readable algorithm, not a model trained on someone's recordings. chiptunes.app github.com/tetrisgm/chiptune… devpost.com/software/chiptun…

Submitted TreeTree to the @OpenAIDevs WebMCP hackathon: the family tree you can talk to. An AI archivist interviews your family into a real, sourced tree. Then your browser's agent drives the page. 14 WebMCP tools, reading, navigating, and creating on the same canvas you're watching. When an agent is driving, the site's own chat bar gets out of the way! Try it with no sign-in at treetree.app/demo, or deploy your family's own with one command. MIT. treetree.app github.com/tetrisgm/treetree

IT'S LIVE! partyparty.party Turn your MacBook into a party. Broadcast music to people's iPhones with no app for them to install.

I’ve come to a realization. Products are either: 1) Built for humans, with AI bolted on as tools 2) Built for agents, with humans reduced to verifiers What if we treated human and non-human intelligence as peers? Humans want UI and chat. Non-human intelligence wants APIs, CLIs and MCP. They’re just interfaces for different kinds of intelligence. In my 2010 TEDx talk, I argued that humanity was already experiencing a slow singularity: peer-to-peer agents collaborating. Non-human intelligence doesn’t change that model. It simply accelerates it! This is the philosophy I’ll be building my next products around. You can find the talk and my product portfolio at ramine.net.

Let me share the biggest lesson about why you want to work with fewer agents and AI automation. I've been vibe coding 12hrs+ a day since May and shipped 4 projects so far. One thing that AI influencers don't mention is that the fancy systems you see here or in Youtube videos about "all the amazing things" agents can do, is technical debt. Creating an agent, or worse, a fleet of agents, means you're sitting with a collection of scripts that will work: 1) the way you currently work 2) only as good as you built the agents 3) with systemic failures It creates a working environment where you have to decide between: 1) fix the current system and feed it further (chasing smaller errors) 2) redo it with new agents (usually better as you learn more and models evolve, but you will be discovering new problems) 3) your work becomes managing compound failures This is often less effective than you doing work directly with Codex or Claude. Your skills files will decay. Your claude.md customization as well. Only automate what you need.

I rebuilt my portfolio to feel like a modern product studio. Over 50 keynotes, projects, publications that I’ve shipped throughout the years. ramine.net

End of an era. I appreciate everyone who spent time with us and tried what we built. We really poured our hearts into it for six years.

Let's take a craftsmanship lens on those 9 games: 1) Breath of the Wild: It knows exactly what it is - recreating the feeling of the original Zelda: adventuring in the woods as a kid. Be a hero. Fight monsters, solve puzzles, explore, cut a tree, climb that mountain. At every beat, there will be something there waiting to delight you.

I think one aspect of AI that resonates with people is how it gives them hope. You can code, design, build a business. Now the onus is on you. It’s very liberating. People feel like they can be lifted. That kind of empowerment is very important when crafting a product.

With the Internet becoming more and more “by AI, for agents”, there is a real possibility that the current web, made of human content and pages, is on its way out. Much like Geocities, or how older consoles are completely backed up and archived (you can download the whole set in a few gigabytes and run it on your computer). It’s quite possible that in 20 years, the human made, public elements of the web are just another platform that is backed up and an interesting moment, before everything became dynamic and generative.

Most people think AI can vibe-code commerce: make a page, throw in Stripe, ship it. Commerce is not a payment button. It is: Fulfillment. Tracking. Email. Contacts. Comments, ratings. Inventory. Access. Support. Trust. The hard part isn’t code. It is making a coherent system.