sanreds

Building @claw_boo (🦞👻) - Open source AI Agent Team Studio. Shipping Daily. Not selling anything, if useful, ⭐️ the GitHub repo↓

[sanreds @sanreds_]

What if you could see your AI agent team?

This is the Ghost Graph from Clawboo - every node is an agent, every edge maps to a real config. Live updates as your team works. Open source. Shipping publicly Tuesday.

[sanreds @sanreds_]

I took the win this week 😎

[sanreds @sanreds_]

Oh the suspense! Will I make it?!

An unspoken war everyone has with @claudeai

[sanreds @sanreds_]

@Luciussus @ClaudeDevs @trq212 I am building clawboo, one platform where all the agents like opeclaw and hermes stay and live together!

https://t.co/aIQQsH8IBa

[sanreds @sanreds_]

@ClaudeDevs @bcherny So far the predictions are holding up 😂

[sanreds @sanreds_]

Claude wisdom take notes:

1. Use up all the week credits with in 24 hrs.
2. Credits will be reset again then use all of it again with in an hour.
3. Then the guy who breaks it again and again breaks it again. 🙃 They will take it out.
4. Anthropic would be sorry and reset credits again for us all.
5. Keep on building tokenmaxxing!!

[sanreds @sanreds_]

@claudeai Weekly credits have been reset.
So far the predictions are holding 😂

https://t.co/EwznqIn9pZ

[sanreds @sanreds_]

Claude wisdom take notes:

1. Use up all the week credits with in 24 hrs.
2. Credits will be reset again then use all of it again with in an hour.
3. Then the guy who breaks it again and again breaks it again. 🙃 They will take it out.
4. Anthropic would be sorry and reset credits again for us all.
5. Keep on building tokenmaxxing!!

[sanreds @sanreds_]

@claudeai @AnthropicAI So far it is true. Now it is at Point 2, the credits have been reset. 😂

[sanreds @sanreds_]

Claude wisdom take notes:

1. Use up all the week credits with in 24 hrs.
2. Credits will be reset again then use all of it again with in an hour.
3. Then the guy who breaks it again and again breaks it again. 🙃 They will take it out.
4. Anthropic would be sorry and reset credits again for us all.
5. Keep on building tokenmaxxing!!

[sanreds @sanreds_]

Not a good way to start weekend.

[Anthropic @AnthropicAI]

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.

Access to all other Claude models is not affected.

We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.

Read our full statement: https://t.co/bwn0sximKZ

[sanreds @sanreds_]

@DhravyaShah @supermemory @openclaw Okay looked at the code. It is the same engine! The hosted enterprise one adds extraction models (proprietary i guess?), connectors and hosted mcp are platform only. Massive!

[sanreds @sanreds_]

@DhravyaShah @supermemory @openclaw You did it just like that 🫢

Is your cloud hosted memory better than local one in any way or is it the inference speed/maintenance/scale you are betting on?

[sanreds @sanreds_]

@trq212 If you had to switch any part of your workflow with opus or lesser models and still get the same output, what would it be?

[sanreds @sanreds_]

@bcherny The “auto switch model” option would save a lot of headache. Opus 4.8 used to stop midway with high false positives if you do anything remotely to security.

[sanreds @sanreds_]

@karpathy @bcherny The “auto switch model” option would save a lot of headache. Opus 4.8 used to stop midway with high false positives if you do anything remotely to security.

[sanreds @sanreds_]

@ViharikaDev Impressive list! Keep them coming 🙌🏼

[sanreds @sanreds_]

Everyone is asking:
“How do we add more agents?”

The better question is:
“Which agents did the new model just make obsolete?”

AI agent architecture is entering its Jenga era. The winners will delete faster than others add.

[sanreds @sanreds_]

https://t.co/WNPMuwAW7G

[sanreds @sanreds_]

https://t.co/WNPMuwAW7G

[sanreds @sanreds_]

Ngl, this AI hardware idea would make people pause. We don’t need another screen shouting for attention. A tiny badge that stays with us and lets us quietly steer our agent squad while moving around feels way more human.

An e-ink version would be even better, calm, low-power, less distracting. More like a quiet companion than another device.

[sanreds @sanreds_]

Co-evolution discipline (Ashby's Law): keep the harness thin where models are improving, defer rigid sprint contracts and verbose role-splitting; keep forever the independent evaluator (self-eval bias is a training property) and structured state/handoff. Re-run the ablation scorecard each major model release; strip scaffolding the new model absorbs.