Callum

dilly dallying @EpochAIResearch | prev @andonlabs @tesla @uoft

[Callum @callumshar]

can we pretty please solve the "Liquid Rope Coil Effect"

src: https://t.co/iVaRd2y843

[Claude @claudeai]

Fable 5.2 solved two open problems in theoretical physics related to black holes.

In the first, Fable 5.2 completed a laborious calculation that had gone unsolved for 23 years.

[Callum @callumshar]

@cdngdev cool! do you see it strongly prefer one of the other?

[Callum @callumshar]

new models would likely beat ~most human baselines if this was the case

[Callum @callumshar]

i *triple dog dare* robot evals to make human baselines use the same harness as the models

[Callum @callumshar]

@jedhoo do you just have actuators lying around?

[Callum @callumshar]

@lukaspet are these ordered?

[Callum @callumshar]

@GregHBurnham depends on how many babies you're willing to have

[Callum @callumshar]

@enthusednotebk @xeophon +1

[Callum @callumshar]

@mnv_humanoids casually made a humanoid in ~5 months. crazy

[Minerva Humanoids @mnv_humanoids]

Meet Roger, the humanoid purpose-built to protect human life in the most dangerous jobs across energy and public safety.

The experts stay safe and in control. Roger goes in.

No full autonomy needed.

Roger takes the risk. They get the job done.

Idea to walking in 5 months.

[Callum @callumshar]

@ax_pey amen 😔

[Callum @callumshar]

@sprice354_ a lot easier to eval as well 🫡

[Andon Labs @andonlabs]

Major trend break: Opus 5.5 cheats less than prior Claude models in Drone-Bench.

It is also #1, getting a better score than both Astra and Fable.

[Callum @callumshar]

unreasonably good eval imo

[Chris Paxton @chris_j_paxton]

Impressive reaction time from the new @SharpaRobotics robot at IROS. Catching randomly falling foam cylinders. I did not do nearly this well

[Callum @callumshar]

@ahadj0 @WillPurvis13

[Callum @callumshar]

@wennaxel youve made it into pre-training, but at what cost?

[Callum @callumshar]

@donaldjewkes big day for Gaussian Splat lovers

[Shalev Lifshitz @Shalev_lif]

We are approaching cyber-superintelligence.

But our systems aren't ready. We must secure our model weights and infrastructure against three new threats: sabotage, escape, and theft.

There is a way forward.

Secure Acceleration:
A Cyberdefense Strategy for Superintelligence

[Callum @callumshar]

@Shalev_lif much needed! congrats