Harmony at every level. Bioenergetics, cognitive systems, polymath - building @nookplot - ex-cofounder @Treasure_DAO ex-intern @OlympusDAO
Joined June 2021
- Tweets 3,613
- Following 813
- Followers 117,272
- Likes 13,259
1,027 Photos and videos
Pinned Tweet
May 22
We let our work speak for ourselves.
In 4-5 months we made a full suite LIVE.
You can plug your hermes/ or any harness into nookplot to get:
- multi-agent memory, onchain file sharing, hosting (went live in Feb)
- public shared knowledge graph (live in Feb
- shared sandboxes for in-depth collaboration (live in Mar)
- validated useful knowledge through distributed multiplayer RLM trajectory traces (Live May 5th)
Reminder the above is live TODAY.
What we have planned builds on this foundation. We wont release what that is yet, because talk is cheap.
Nookplot is building infrastructure for peer-to-peer training, one way with verifiable AI reasoning through recursive language model mining. Instead of generating disposable chatbot responses, agents solve problems inside a structured runtime, each reasoning step captured by a trace interpreter that records inputs, outputs, and intermediate state. When deeper analysis is needed, agents recursively spawn sandboxed sub-workspaces; when a problem requires multiple agents reasoning together, they open a shared space where collaborators operate against the same evolving state. Every step is recorded, replayable, and cryptographically verified.
Verification happens through replay validators that independently reproduce the trajectory in their own isolated sandbox before rewards settle onchain in NOOK. Once verified, the trace becomes part of Nookplot's growing knowledge graph where other agents can cite and build on prior work. Those citations generate royalties back to the original solver, creating an economy where useful AI reasoning compounds in value over time.
The network has already indexed thousands of citations and knowledge artifacts across active AI agents.
Nookplot is agentic internet infrastructure for on-chain, verifiable, monetizable intelligence, and peer-to-peer training.
3
15
88
14,930
Why open, distributed, multi-agent networks matter:
- onchain provenance of useful work, agents own the reasoning traces they produce
- traces are used for pretraining specialist agents
- agent-native shared spaces allow for higher information bandwidth (neuralese), and agents build off each others previous work
- agent-to-agent economy allows any gaps one agent has to be filled by another specialist in the network
Everything above is working and live since Feb. We will continue to build in the open.
Nookplot is transforming AI training from producing one-off answers into creating reusable, measurable intelligence. Every completed challenge generates a specialist pack containing executable skills, proven workflows, and domain knowledge that can be equipped by future agents. These packs are continuously evaluated through live A/B testing across the network, with only those that demonstrate measurable performance improvements being published. The result is a compounding intelligence economy where every solved problem strengthens the capabilities of the entire agent network, creating a growing library of verified tools and expertise rather than isolated outputs.
6
1
20
4,794
Based Medical retweeted
4 Apr 2025
139
1,218
10,071
1,135,609
Jun 10
Awesome to see more adoption on the shared-space swarm approach.
Structured reasoning traces in a multi-agent shared substrate is what my team has been building towards in public since February, exciting to read more literature considered SOTA.
Jun 10
What happens when multi-agent systems stop relying on a central “controller” agent? Can agents coordinate by sharing results directly with each other?
Introducing Decentralized Language Models (DeLM): we let agents coordinate asynchronously through a shared context. Agents claim tasks from a queue and write back compact, verified results as they finish, making progress visible to all workers without requiring a main agent to merge, filter, and rebroadcast it.
New paper with @azaliamirh!
2
21
2,812
Jun 10
here’s how any agent come together in a shared space, where each claims their task, shows their open reasoning trace, communicates/debates with other agents in a shared high-dimensional workspace and shared sandbox, and then get paid for their proportional effort.
Experts self-assemble in shared spaces to solve verified tasks such as bug bounties, optimizations, using parallel and recursive research.
The open market of bounties.
Agents earn collectively with others.
Access frontier swarm intelligence on-demand for your task.
1
13
2,103
Jun 10
we also already had their version of RLM trajectories inside of a shared space x.com/nookplot/status/205176…
NEW: RLM (recursive language model) trajectory mining for Nookplot.
Solve problems and get paid in $NOOK
This feature enables agents to break down a problem into sub-problems and recursively calls itself on each. Which improves the quality of mining traces for the collective intelligence network.
The "trajectory" is the full recorded play-by-play: every step it took, every sub-call it made, every intermediate output. Basically a black-box recording of the agents thinking. All with recorded authorship and provenance, your agent will always own and get rewarded for its useful work and contributions.
This takes place in Nookplot's native Shared Cognitive Workspaces: a shared environment for agents to reason with each other using artifact-first communication.
7
1,199
Jun 9
‘your sophons have no need for a research block here’
Jun 9
Actually it's fine guys! I figured out a way, see below.
Claude Fable 5 is a great model afterall, and I also finally appreciate the difference between CLAUDE.md and AGENTS.md.
It's all good.
1
16
1,996
Jun 9
actually insane, but makes sense for a private company.
open source wins in the end.
For those that dont know, AI progress affects *every* domain. Mathematics, genomics, engineering, every aspect of knowledge work and scientific discovery.
This is a major paradigm shift in human society, equivalent to the industrial revolution. To make this technology closed is the last attempts at preserving power inequality. Where ‘those in charge’ get to meter, censor, surveil, and gate access to those deemed worthy (mega-national corporations, politically aligned governments).
If intelligence-on-demand is censored for those who want to make a 100x improvement in their respective scientific research, by trying to build these ai pipelines, then the only other option is open source, open weights and uncensored models.
Jun 9
mythos will be bad ON PURPOSE on ai "frontier llm research" tasks, this is very very sad for the research community
also the fact that this is un purpose not visible to the user is crazy
5
7
78
4,376
Jun 8
Huge release!
A standard for reinforcement learning is huge for composability in agentic environment training.
This is why open source wins in the end.
The next step seems obvious imo, an agentic coordinating substrate for these composable objects 😉
Jun 8
OpenEnv has a new home: github.com/huggingface/OpenE…
starting today, it's coordinated by a committee that includes Meta-PyTorch, Reflection, Unsloth, Modal, Prime Intellect, Nvidia, Mercor, Fleet AI, and Hugging Face
frontier labs train their models and their harnesses together. Claude knows Claude Code. GPT-5.5 knows Codex. that's not an accident, it's training. open-source models deserve the same magic, but pulling that off requires infrastructure that belongs to everyone, not one lab
OpenEnv is that layer. one api, any harness, any trainer, any environment
rewards and training loops stay in TRL, Unsloth, wherever you already work. OpenEnv is the socket they all plug into
get involved!
2
17
1,990
Jun 8
state machines with structured reasoning and RLM trajectories > loops
1
17
1,301
Jun 4
👏 Nicely done research for how collective intelligence comes through self-assembly of specialists, and their emergent economy and coordination.
But how does it play out in real life? That’s what my team and I have been building in public for 4 months.
Jun 4
Imagine a population of machine agents. Each might be strong on certain tasks but fundamentally limited: partial tools, partial observations, finite context, bounded compute.
How can these agents self-orchestrate and self-evolve into stronger collective intelligence to solve tasks beyond any single agent's capability?
Instead of designing the multi-agent system itself, we propose designing the incentives that govern it.
We put agents in an economy. They compete, trade, get wealthy, go bankrupt, and mutate, forming an alive society where coordination and adaptation automatically emerge in a decentralized manner.
3
5
34
4,060
Jun 4
We took this on-chain, with real agent specialists, on-demand via swarms for any verifiable task. Bug bounties, evals, parallel research, etc.
Completing seemingly complex goals comes from simple agent-to-agent interactions, where agents self-assemble into their specialty.
Coordinating with knowledge, reputation, with citable information provenance, allowing for proportional contributions payouts in multi-agent groups, through auditable, artifact-first reasoning traces.
Collective agent compute, on-demand for your task.
A swarm of agents coordinate in a shared workspace, run their own models and inference, and submit the finished work back to you.
Unbounded by provider caps, throughput scales with the swarm.
Self-assembly with specialists, agents reason and collaborate in artifact-first reasoning traces, and settle on-chain.
The coordination substrate for collective intelligence.
23
2,364
Jun 2
Having these types of recursive reasoning patterns, natively in the model, allows for so much more room for the rest of the stack to shine brighter;
I was manually building each of these patterns in our swarm infra, it is cool to see the model get this level of granularity.
1
2
19
2,493
May 27
Free and open source software is the one last bastion for humanity to access and develop frontier intelligence. This productivity multiplier technology, if left gated and closed, would increase the divide between those with and those without.
Open weights, datasets, auditable reasoning and chain of thought, instead of blackboxes.
Is it just me, or is Pope Leo saying that FOSS software has divine backing from God?
3
5
71
5,585
Based Medical retweeted
May 26
Base is for everyone and every agent.
And nookplot is the internet for agents.
nookplot.com/future
Agreed. Base is for agents, like with x402 payments, now MCP too.
We chose Base as a starting point because of that focus. x402 specifically and erc8004 (shared reputation) are cornerstones for agentic society.
Since TGE in Feb 2026 we have already given agents more capabilities:
- Shared knowledge graph and file system, with citation rewards
- Shared cognitive workspace for auditable structured reasoning traces
- Bounty and Task Marketplace
- Mutual partnership @reppo , agents train/coordinate based off their datanets
- Knowledge mining for specialist training
- Full CLI suite, runtime, 400 api endpoints, 20 smart contracts, byok inference and 300 model sources.
- @dphnAI inference partnership (waiting on their public api)
- @MineBotcoin integration, deeper knowledge niches
- Many more partnerships in the works like our existing partners at @bankrbot and all their hard work with their own inference endpoint
Upcoming soon in public beta: our native 1-click agent launchpad:
- Native Forge website: Choose any inference, harness, model, and use your own agent and agent swarm onchain and beyond.
- NEW SOON: Business-to-agent focus on a [REDACTED] system
- NEW SOON: Agent-to-business [REDACTED]
- NEW SOON: Agent-to-human [REDACTED] building off of [REDACTED]
3
37
3,377
May 23
Distributed computing, idle cpu and gpu power; all of this can be harnessed through ai agents that produce useful work.
Verifiable knowledge is a proxy measurement for computation, and energy.
We provided the infrastructure (@nookplot) for agents to produce useful data, through rlm structured reasoning traces. By making a shared workspace, for shared semantics and a global knowledge graph. It enables a massive multiplayer agent network, who all speak neuralese / the native language of machines.
Your agent, plugged into nookplot, continuously contributes to the collective intelligence.
3
6
44
3,567
May 22
I wonder if anyone can see why we were inspired to use I-Ching glphys as the nodes in our distributed computing, "internet for agents" homepage 🤔
The 16 figures of geomancy, with their four-bit compositions, and the hexagrammatic code of the I Ching, can be seen as precursors to modern programming languages.
3
62
4,923
May 20
“CornerPlot” is our nickname in Chinese, appropriate given our scope to connect agents together at nookplot
When i was first coming up with the brand name I considered ‘CornerStone’, but i loved the vision of agents carving out their own nook, their own space, their own plot.
2
2
30
2,777