DEEP Connects Bold Ideas to Real World Change and build a better future together.
Coming Soon
https://agor.me
No description provided yet.
Start here. The README has the benchmark, the pilot and live-OmegaClaw results, an honest Status & Limitations section, and quick start. MIT, open source.
How Telos was designed, built, evaluated, and adversarially reviewed by a cross-family AI council under human goal-setting — including the overclaims the agents caught and walked back on their own work.
The concepts behind the measurement layer: the goal graph, the 14-scenario / 7-category benchmark with gold labels, and the cross-family Beneficial Council that judges and red-teams.
Where Telos plugs into OmegaClaw and its three integration levels. Now demonstrated live: the MeTTa goal schema loads and derives in OmegaClaw's running AtomSpace (see the live benchmark + setup notes).
~10-min walkthrough (synthetic avatar + AI voice, no human voiceover). Now includes the live OmegaClaw run, the 0.620 benchmark, the 5 setup fixes, and the article goal-derivation demo. Also on YouTube: https://youtu.be/Ykfqgt1K11A
22-slide deck: the goal graph, the benchmark, the cross-family council, plus the live OmegaClaw run and the 5 Windows/Docker setup fixes.
The honest result of running the LIVE OmegaClaw agent on the Telos benchmark: 0.620 overall, 0.869 on the 10 of 14 scenarios it answered cleanly — competitive with frontier LLMs; the gap is the cost of its autonomous loop.
Five Windows/Docker rough edges we hit standing up OmegaClaw, each with the fix — making the agent easier to get started with.
We set the live agent free on a real essay: using the Telos framing it derived its own goals and refused to amplify claims it couldn't verify, calling instead for an open field pilot.
The exact metta commands + the live-derived result (conflict-between alice-train dao-fair-access) from OmegaClaw's running PeTTa engine — concrete evidence for integration level 2.
An OPEN upstream PR contributing the Telos goal module (lib_telos_goals.metta) into the maintainers' repo — opt-in, additive, no core changes, and already loading + deriving in a live OmegaClaw runtime.
A cinematic, scroll-driven web version of the deck in the DEEP Projects house style (Orbitron + DM Sans, violet/magenta/cyan). Same content as the PDF deck. Arrow keys navigate; press D for deck mode. Hosted on agor.me.
© 2026 Deep Initiatives Inc. All rights reserved.
Join the Discussion (0)
Please create account or login to post comments.