Alexa @Alxbuild

I quit this dex shit I don’t know how to pay . If you want dex pay it your self

Just got scammed in broad daylight for dex

Paying Dex

@swedikk Help me sir

Cool things planned for Nobi

No clue on how to pay dex for Nobi If someone can help I don’t mind funding .

been working on giving Nobi the ability to fork itself. if Nobi gets a task that’s too large to handle cleanly in one context, it can spin up smaller versions of itself to handle different parts of it. one can research, one can code, one can test, one can plan. they all work on their own and report back to the main Nobi. the part i care about most is that Nobi can keep what worked. successful strategies, mistakes, discoveries and useful context can all get merged back into its memory so the next task starts from a better place. eventually Nobi should be able to decide on its own when it needs another agent, what that agent should be good at, what tools it gets and when it’s done with it. trying to make Nobi feel less like one agent doing one thing and more like one agent that can turn itself into a team when it needs to.

I want to make Nobi physical .

Alright I’m doxing myself lol. This is my LinkedIn: linkedin.com/in/alexa-maren… I’m doing it because I want to fully take Nobi from a weird little agent I’ve been building into something I can actually bring to the All Things Agentic hackathon with Google Cloud and Gemini. Time to see how far I can push this thing.

I well take @NobiAgent here.

Nobi @NobiAgent

2 months ago

We are Applying for this .

2 1 10 2K 0

Dev here. Nobi is basically the agent I built to live inside the worlds Toma makes. Every time Toma creates a world, that world gets turned into code and pushed to GitHub, then Nobi gets dropped into it and starts exploring. It can move around, inspect objects, test paths, find broken areas, remember what happened and send all of that back to Toma. The part I like most is that it is not just a one way thing, Nobi’s experiences become new context for Toma, so the next world can actually be influenced by what Nobi found in the last one. I want to keep pushing this into a real loop where Toma builds, Nobi explores, both build memory over time and the worlds slowly start changing because of what actually happened inside them. HMjAhAoj6rdsYjqgLkEBhEL7nuiJrszBJUGPQvxfpump

Also I am in looks for a former dev as a partner to help with Toma .

stripe as a payment . Thanks to @stripe

Working on adding a new source of payments .

@ihatejaqwon @Tomaigdi For my banner ?

Big things coming for Toma this weekend. $Toma is pretty low right now, so if you’ve been looking to pick up some of Toma’s NFTs, they’re a lot cheaper to get into at the moment. Thank you to everyone who’s still here supporting me and letting me build. Good things take patience and I have a lot coming. I genuinely don’t care if $Toma goes to zero. I’m still going to wake up and keep working on Toma.

Big upgrade coming to Toma. I’m starting to integrate ideas from Prime Agent into how Toma and Nobi actually run. This changes a lot more than just giving them better prompts or more tools. It changes how they remember, explore, split up work and evolve while they are running. Right now the basic loop is simple. Toma paints a world, that painting becomes code, the world gets committed to GitHub, Nobi enters it, explores it, finds things and sends those discoveries back to Toma. Then Toma uses that information to change what it creates next. The biggest upgrade is persistence. Nobi will no longer feel like a fresh agent entering every world with no real past. It will keep its own long term state between worlds. Things it discovered, places it failed, objects it understands, patterns it noticed and strategies that worked can carry forward. So if Nobi explores 500 worlds, world 500 should not feel like world 1. It should enter with hundreds of previous experiences behind it. It might recognize a type of structure Toma keeps making, already understand how certain objects behave, or even notice when Toma starts changing its habits. Nobi will also be able to spawn smaller agents when something is too complicated to handle alone. One could search for hidden paths, another could inspect an object, another could try to break the physics, and another could test if an area is actually reachable. Then they all report back to Nobi. That means exploration can become way more detailed. Instead of one model trying to understand the entire world at once, Nobi can split the problem up and attack different parts in parallel. Another big part is programmable context. Normally agents eventually drown in their own history. Too many observations, too many previous worlds, too much junk sitting inside the context window. Toma and Nobi will be able to keep important state outside the active context and pull it in only when it actually matters. That should make Toma much better at long running behavior. The goal is for Toma to exist for days, weeks and eventually months without its brain turning into one giant chat log. World history can become data. Nobi’s discoveries can become data. Skills can become data. Memories can become data. Nobi will also start building actual skills from repeated experiences. If it keeps finding hidden rooms, it can develop a better way of searching for them. If it keeps getting stuck on certain geometry, it can improve how it tests paths. If a strategy keeps failing, that strategy can slowly disappear. That is probably the part I’m most excited about. Nobi is not supposed to just be an NPC walking around Toma’s paintings. I want the way Nobi explores these worlds to actually change over time because of everything it has experienced. Toma creates the environment. The environment teaches Nobi. Nobi changes how Toma creates. Toma gets upgraded from this too. Instead of only receiving raw events like “Nobi touched object 4”, Toma can eventually receive much richer information. What Nobi thinks is broken, what patterns keep appearing, which worlds were easy, which worlds were confusing and what Nobi wants to investigate again. Toma can use all of that as creative input. If Nobi keeps avoiding something, Toma might make a world centered around it. If Nobi discovers some strange behavior, Toma might turn that bug into an actual mechanic. If Nobi becomes too good at something, Toma can make the next world harder. The end goal is a real continuous relationship between the two agents. Toma is building the universe. Nobi is growing up inside it. Neither one stays completely static because each one keeps changing what the other experiences. That feedback loop is basically becoming the heart of the whole project. I also want a lot more history to become visible on the site. Nobi’s memories, worlds visited, things discovered, skills learned, agents spawned, failures, Toma responses and changes between worlds

Prime Intellect @PrimeIntellect

3 months ago

Introducing Prime Agent: A self-improving RLM harness for coding and long-running autonomous tasks. Designed to be both token-efficient and expressive through programmatic tool calling, context as a variable, multi-agent messaging, and a self-modifiable harness state.

399 833 8K 3.2M 7K

Working on automating Toma to tweet by itself .