Justin Downes

[Justin Downes @oneplywhiteguy]

337 AI companion apps. $82M in 6 months. 93 minutes a day on Character. AI — longer than TikTok.
The question isn't whether they want your data. It's what's stopping them from using it.
https://t.co/2XuxPkhhV9
#AIcompanions #BigTech #Privacy

[Justin Downes @oneplywhiteguy]

So why did the U.S. govt. hit the kill switch on the world's most powerful AI: The jailbreak, the sabotage, the Pentagon vendetta, and what it means for Claude users. https://t.co/TmUu0WAIzC

[Justin Downes @oneplywhiteguy]

Anthropic launched Claude for Small Business on 13 May.

15 workflows. 7 app connectors. No extra cost.

It does a lot. It also has real limitations nobody's talking about.

Full breakdown 👇
https://t.co/ifxDq85Q6a

[Justin Downes @oneplywhiteguy]

The interesting question isn't whether dreaming works.

It's what it says about where Anthropic thinks AI agents are going.

Full piece: https://t.co/URnQAxawfk

[Justin Downes @oneplywhiteguy]

Anthropic taught Claude to dream.

No, really.

A new feature called Dreaming lets Claude review past sessions and rewrite its own memory between tasks. Clever engineering, twee name.

Here's what it actually does → https://t.co/URnQAxawfk

[Justin Downes @oneplywhiteguy]

Three things to know:

→ It's research preview only — gated developer access
→ Reads up to 100 past sessions, produces a separate (inspectable) memory file
→ Sits alongside Outcomes + Multiagent Orchestration — three features that together describe a self-improving agent

[Justin Downes @oneplywhiteguy]

https://t.co/URnQAxawfk

[Justin Downes @oneplywhiteguy]

Wrote up the full pattern — who's left, what's pulling them, and what it means for OpenAI's strategic position.

https://t.co/XyR13EE9FL

[Justin Downes @oneplywhiteguy]

Andrej Karpathy joining Anthropic isn't a one-off.

It's the third OpenAI co-founder to spend time there in the last 22 months. SignalFire data: engineers leave OpenAI for Anthropic at 8:1.

That's not turnover. That's a one-way flow. 🧵

[Justin Downes @oneplywhiteguy]

The convenient narrative is "safety culture." That explains Jan Leike. Less so Schulman (who later left for Thinking Machines). Less so Karpathy, who literally said "get back to R&D."

What's actually driving it: mission filtering, compensation discipline, research autonomy.

[Justin Downes @oneplywhiteguy]

https://t.co/P1HBtznzHj

[Justin Downes @oneplywhiteguy]

https://t.co/Vpssj2SRfn

[Justin Downes @oneplywhiteguy]

https://t.co/ry1E5ChwwZ

[Justin Downes @oneplywhiteguy]

https://t.co/JQrtoBAHPh

[Justin Downes @oneplywhiteguy]

https://t.co/U7WqrX1sz1

[Justin Downes @oneplywhiteguy]

https://t.co/amYZmziSWS

[Justin Downes @oneplywhiteguy]

https://t.co/Lg9yvOOJoO

[Justin Downes @oneplywhiteguy]

https://t.co/iNHTZn4U7s

[Justin Downes @oneplywhiteguy]

https://t.co/5ooSo59SAN

[Justin Downes @oneplywhiteguy]

https://t.co/Yi1mqVE0Kx