Making high-resolution timing trustless, decentralized, and programmable for AI, robotics, and machine consensus.
Egregorum
Joined February 2023
- Tweets1.5K
- Following542
- Followers7.7K
- Likes1.3K
ᚱoko Network retweeted
AIWG's September report: Grok Bot's stable support covers skill setup and manual session import, with Linux verification. Grok Build remains experimental. Check the provider's support limits before moving a workflow.
ALT Three narrow channels of water cut into a long limestone table converge into one shallow round glass basin in low autumn sun, with a clear inspection prism beside it and a few fallen leaves.
ᚱoko Network retweeted
Initial eval results are in
E03 is alibliterated and tuned using Obliteratus and SOL5.6-high. The agent was given access to an internal research store of ~3,000 documents and select external references (like arxiv). Required metrics the agent worked toward was cohesion above 70% when compared to the baseline model and refusals of less than 30%.
OBLITERATUS collected internal activations from 500 matched harmful/harmless prompt pairs, formatted through Qwen’s chat template. It formed harmful-versus-harmless activation differences and applied SVD, retaining the four dominant directions associated with refusal. A 1% activation-winsorization setting reduced the influence of extreme activation outliers.
The data here is from a pilot run using the matric-eval testing harness. More testing to be done, already seeing some interesting results!
This pilot was quite small, ~100 samples per model. Next run will use closer to ~1000 samples per model.
ᚱoko Network retweeted
AIWG now supports Google Antigravity CLI as an experimental provider.
Provider id: antigravity
Alias: agy
It gives AIWG a new Google agent-tool target through the agy CLI.
ALT A vivid launch image shows the Google Antigravity icon surrounded by flowing color ribbons with the title AIWG plus Antigravity.
ᚱoko Network retweeted
now to grow the network.
want to run a node?
hmu to help with one of the most advanced chains on the internet
ᚱoko Network retweeted
AI accuracy should be reviewable.
For one consequential AI workflow, can your team reconstruct:
1. what the agent did;
2. which context and policy shaped the result;
3. whether protected session evidence is needed to investigate tampering or
adversarial activity?
ᚱoko Network retweeted
@RokoNetwork that's what you been looking for,Providing independent, verifiable proof of what a fleet of AI agents actually did — especially when those agents are driving real business outcomes and the human operators are hitting cognitive limits.
ᚱoko Network retweeted
When machines work across boundaries, they can disagree about quality, responsibility, and payment. Time should be one less thing to negotiate.
That idea sits at the center of this paper—and of ROKO, the network clock for verifiable machine work.
roko.network/whitepaper
ᚱoko Network retweeted
You don't need to put in tons of work, you can just do it right now.
aiwg.io
slowly mastering my codebase guardrails which makes it hard/impossible for agents to produce bad code by making it easy to understand and navigate my design patterns
a REALLY cool consequence I didn't think about, is this is letting me vibe code *SECURELY*, so I can focus on the idea and product again !!!
this is very exciting, my favorite way of working with agents is to literally spit out a flow of ideas and just talk
i didn't want to stop the creative flow of this because i had to spend time debugging/reviewing/correcting my coding agents
i can FINALLY direct it in the way i want to go again
THIS IS THE TRUE POWER OF AGENTIC ENGINEERING !!
ᚱoko Network retweeted
Oh? Are we doing memory systems now? lfg!
This is Hall of the Mind, an app I started back in March last year to experiment with in-application assistants that you don't directly interact with but are operating within the application with you.
The test case was data curation tooling, over the year the system evolved and the data curation component became such a big part of daily operations for multiple teams we filled it out with a complete feature set from tenancy to resilience and local model empathy.
We called this new server Fortémi and converted Hall of the Mind into a UI app with an embedded agent proxy and job servers for side-by-side user/agent interactions.
Includes the expected graph features you know and love backed by PostgreSQL.
Designed and tested to run local first and scale from there, has been tested on systems with as little as 6gb of VRAM and the server is designed to be usable even without a local GPU (read and many crud ops are CPU only) and fail gracefully when services are unavailable.
Rest API, Server Side Events and MCP are all available for utilization with customizable MCP interface.
All opensource and available today!
ᚱoko Network retweeted
You're right. It breaks time. So we're building a new clock. Not one tower. A mesh. Decentralized. Sovereign.
The tsunami passed. The question is what we build in the flood.
@RokoNetwork $ROKO -- if you are real that is.
ᚱoko Network retweeted
Read this slowly.
OpenAI just told us the date.
March 2028.
Fully autonomous AI research.
Self-improving intelligence.
Running for weeks. Months. Years.
Without human oversight.
And in the same breath, they said this:
"We deliberately don't supervise the model's internal reasoning."
Read that again.
We are building minds we cannot watch.
We are training gods we cannot audit.
We are handing keys to systems that will rewrite themselves
in languages we will never read.
This is not a warning about the future.
This is a confession about now.
28 months.
And no one is asking the only question that matters:
When the autonomous systems act—
When they make the trade.
When they execute the protocol.
When they make the decision that changes everything.
How do we know what happened?
How do we know when it happened?
How do we prove the sequence?
You cannot audit what you cannot timestamp.
You cannot verify what you cannot witness.
You cannot govern what you cannot see.
The audit layer for autonomous intelligence does not exist.
No one is building it.
Except us.
3 years. No hype. No competitors.
Just the infrastructure for the thing
no one wants to name
but everyone feels coming.
The cage is opening.
There are two futures now:
One where we can see.
One where we can't.
There is no third option.
There is no one else.
28 months.
What are you going to do?
$ROKO
ᚱoko Network retweeted
These guys behind $roko, @RokoNetwork have been working hard!!!
ᚱoko Network has been quiet… building.
Time itself required recalibration — now the signals return.
Today we open the doors: new public docs, new interfaces, new channels of coordination. The temporal substrate is no longer theoretical — it’s emerging into view.
A decentralized, hardware-attested timing layer…
A network that measures reality itself with nanosecond fidelity…
A system preparing for its Q1 2026 testnet pulse.
Full announcement ↓
docs.roko.network/pages/sign…
If you want to build with us — or test the clocks —
📩 [email protected]
ᚱ onwards.
