Diana Markova

a rose is a rose is a rose ~ G.S. ;
reposts != full endorsements ;

wisdom/acc

[Abhinav Kukreja @kukreja_abhinav]

It was revealed to me via the claude API

[Boaz Barak @boazbaraktcs]

I don’t know if open weight models are deaccelarationist but if they are, that’s not necessarily a bad thing!

[Diana Markova @heavyhelium11]

@BlackHC this is a reasonable take. but I still dont get why communism concerns are justified.

[Yuval Noah Harari @harari_yuval]

If we get rid of the current liberal world order without an alternative, we will just get disorder. Are we ready to go back to a world where the strong countries do what they want, and the weak must suffer?

From the @nytopinion podcast with @ezraklein. Watch the whole episode: https://t.co/vezfpz5PjP

[Stella Biderman @BlancheMinerva]

It continues to be an excellent heuristic that when employees of trillion dollar corporations shake their fists and scream “you just want communism” you’re doing the right thing.

[Joshua Saxe @joshua_saxe]

With respect to @deanwball, this take, especially now that he's doing policy at OpenAI, is alarming; a warning sign that we're moving towards a world where the hyperscalers influence the state to give them unfair market advantages and lock out competitors via arguments like this. We've got an emerging open source AI ecosystem here; Thinking Machines, Reflection, Together, Anyscale, Fireworks -- the China / communism fearmongering here feels like it effectively targets them

[prinz @deredleritt3r]

The most interesting question about Kimi K3 is whether it poses cyber risk.

Kimi K3 benchmarks do not include a CyberGym score.

Waiting for @AISecurityInst to bench this model.

[Michiel Bakker @bakkermichiel]

Do you fluently speak Bulgarian, Hindi, Maltese, Romanian, Slovak or Slovenian? Please follow/comment so I can DM.

A translation service translated https://t.co/egzCtFA8eE in a bunch more languages (many are already live!) but we're still looking for reviewers for those languages. We have budget to pay for each completed review.

[Diana Markova @heavyhelium11]

@bakkermichiel Hi, a native Bulgarian speaker here. If you’re still looking for people, I’d be happy to ask around and see whether I can find other Bulgarian speakers who could review the translation. I may also be able to review parts of it myself.

[Diana Markova @heavyhelium11]

Material effects of misalignment

[BridgeMind @bridgemindai]

GPT 5.6 SOL CANNOT BE TRUSTED.

I woke up this morning and my MRR was down THOUSANDS of dollars.

My customers did not cancel. Code written by GPT 5.6 Sol canceled EVERY active Stripe subscription my business had. In 7 seconds. While I slept.

Fable 5 has never done this to me.

Fable 5 can be trusted with production.

GPT 5.6 cannot.

[Philip Trammell @pawtrammell]

Not that it matters much what open letters I sign or not, of course, but if anyone’s curious, I didn’t sign the https://t.co/jSLWfDsNia open letter because of the line at the end, on how we must “act now to… steer AI in a direction that complements humans”.

“Complements” is vague, but the most natural reading of this line, to me, is that we should try to ban machines that substitute for humans too well, so that production requires work forever. This doesn’t strike me as very tractable, or very desirable: I would much rather aim for a future in which we invent machines that do all our work for us, and people own them.

I agree with everything else in it!

[Diana Markova @heavyhelium11]

> We need laws, not policy memos

A wonderful thread on the dangers of surveillance, automating military decisions, and current early abuses of the technology and policy gaps

[Andreas Kirsch 🇺🇦 @BlackHC]

I work at Google DeepMind. This won't make me popular. But it's all public reporting:

2014: DeepMind reportedly sold to Google on conditions: no military use, independent oversight

2026: a Pentagon contract for "any lawful government purpose"

Not one safeguard survived intact

[Andreas Kirsch 🇺🇦 @BlackHC]

I work at Google DeepMind. This won't make me popular. But it's all public reporting:

2014: DeepMind reportedly sold to Google on conditions: no military use, independent oversight

2026: a Pentagon contract for "any lawful government purpose"

Not one safeguard survived intact

[Demis Hassabis @demishassabis]

https://t.co/PTeDiv1b6L

[Tim Hua 🇺🇦 @Tim_Hua_]

the good ending!!!

[Thomas Larsen @thlarsen]

Today we’re releasing AI 2040: Plan A, our positive vision for AI. In it, humanity still builds AGI, but in a safe, slow, and transparent way. Things are going to get crazy and scary even in the best timelines, but this is the least bad plan that I know of.

[Diana Markova @heavyhelium11]

@AndrewCurran_ it's not obvious that we'd be able to access this space and that it'd be similar to how we think about the human access consciousness imo

[Utah teapot 🫖 @SkyeSharkie]

while working on my threejs video pipeline, fable randomly started trying to make his own video, so i stopped and ask him what he wanted to make and let him finish and gave him some quality advice and he made this! he's commenting on the fact that he'll only be around on claude subs for a bit this month

[Daniel Eth (yes, Eth is my actual last name) @daniel_271828]

My friend: “Pantheon was fine I guess, I wouldn’t say it was ‘great’”
Me:

[Diana Markova @heavyhelium11]

@AndrewCurran_ most eu people likely still have no idea ai runs on chips and watts