Wyrd

Escaping inherited realities

[Wyrd @wyrdmaze]

@gautham_city_ Genius marketing by @NotionHQ !

[Wyrd @wyrdmaze]

@thesupermannx It's finally happening!

[Wyrd @wyrdmaze]

Welp…

[Cohere @cohere]

While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.

Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.

Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.

But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?

We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.

Read the full report: https://t.co/TZSugXi2ZE

[Cohere @cohere]

While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.

Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.

Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.

But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?

We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.

Read the full report: https://t.co/TZSugXi2ZE

[Wyrd @wyrdmaze]

@bneyshabur Super impressive but is context window scaling the right thing to be focused on?

[Behnam Neyshabur @bneyshabur]

Introducing Fern, Mirendil’s first model: pretrained from scratch, with up to 10M tokens of context!

That’s ~0.5M lines of code (roughly Megatron-LM scale) or days of an agent’s work, all in context at once.

Its first version, Fern 0.1, had a short round of post-training 🚢

[Wyrd @wyrdmaze]

The Spotlight Effect: a psychological cognitive bias where people overestimate how much their appearance, actions, and mistakes are noticed and judged by others

[Wyrd @wyrdmaze]

@waifupika @BrendanFoody mercor was 2nd to scale

[Wyrd @wyrdmaze]

#̡̡̢̡̡̧̡̢̨̡̛̛̛̛̳̪̰̩̟̮̪̣̫̙̠̱̘̜̭̙̥̙̖̬̫̘̖̩̫̥̭̗̭̬̞̥̮̖̖̬ͨͣ̐̆ͨ̎̏̃ͧ̑ͤ́ͧ̄ͦ̇̏ͦͩͣͩ̔̈̎̊̈̋̀̄̒̆ͦ̂́̈̃̑ͨͧͨ̆̓ͬ̌̄̎̆

[Wyrd @wyrdmaze]

It's scary to think how far the education system will be in a few years

[kache @yacineMTB]

My moat is that I am retarded and have no shame. Which means I will do the retarded thing that works

[will brown @willcb]

“continual learning in the harness” is like if a computer had a cache and a disk but no RAM. like sure it could technically work but it’s clearly not optimal. you want RAM.

[Wyrd @wyrdmaze]

@claudeai Are we just afraid to try different things???

[Wyrd @wyrdmaze]

@askalphaxiv There’s got to be a way to do selective forgetting

[Wyrd @wyrdmaze]

Can we normalize demos that make adults feel like kids again? 🚀🚀🚀

[Tesla @Tesla]

Do a kickflip!

[Wyrd @wyrdmaze]

@Tesla Teslamborghini

[Wyrd @wyrdmaze]

@Tesla Demo so good that we all thought it was AI generated

[Wyrd @wyrdmaze]

@yar_vol @paulg Go check out their website. They have a demo. Definitely not a scam.