Umar retweeted
Will that mean keeping the Cybertruck in production beyond next year?
🚨 JUST IN: Rep. Carlos Giménez has formally invited Elon Musk to RELOCATE Tesla's headquarters from Texas to FLORIDA!
Giménez cited concerns over newly elected Democratic Socialists in Texas, arguing Florida offers a more business-friendly environment.
“Florida is rolling out the red carpet for innovation, investment, and job creation.”
This actually makes me dislike the crybercab idea, only because Tesla think the right move is to take away products customers can buy and instead try to do the only think that leads to most profit.
We are still building Cybertruck for another 8 or 9 months btw, but then Tesla needs to focus on scaling Cybercab.
There will never be another vehicle like it.
41
What happens when you stop making products? @elonmusk
Tesla is going from being the most exciting company to maximizing profits over anything. What happened to “the future should look like the future”. We can’t buy the cybercab, and frankly who is going to buy a 2 seater? There are known plans for a bigger family size cars. And now they’re discontinuing possibly one of their most exciting products, the Cybertruck to go all in on autonomy as well.
What is left for consumers to buy? 3, Y, roadster? Extremely disappointing.
28
I’ve seen new model drop predictions every day for the last month. Where are the new models?
Opus 5.5 still goat
20
Umar retweeted
Background subagents in Claude Code now use less of your usage limit.
A running subagent updates its status line every 30 seconds, and each update used to re-read its whole conversation. Now Haiku 5.5 writes it from short notes instead, so each update costs about 99% less.
I’ve hardly used codex today after the reset and my weekly limit is already down to 40%.
I’ve been working all day on opus 5.5 and have hardly had to worry about the limits.
2
135
Wild times ahead
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: cohere.com/blog/concerning-b…
20
