low is the way to the upper bright world
Joined August 2009
- Tweets 8,094
- Following 2,148
- Followers 2,249
- Likes 72,252
938 Photos and videos
there will be a lot less rallying around anthropic this time around. instead research and industry will derisk. the models arenβt worth the volatility / data retention / subtle sabotage over inscrutable and dubious safety standards. they squandered trust
22h
If you needed proof whatβs happening is pure politics and not actually based on the model capabilities, here you go:
βAfter the meeting, Anthropic CEO Dario Amodei and administration officials spoke on Friday, said four people familiar with the call. During the conversations, Anthropic officials laid out how the security vulnerabilities found through the alleged jailbreak were relatively simple and could be achieved with other models. But the government told Anthropic that it had already decided to implement the export control.β
3
159
darren retweeted
1/
What if you could train a model on totally benign-looking Wikipedia articles, but secretly force its internal weights to encode a fully functional QR code?
This is now possible. We can program neural network weights using natural words.
π§΅
2
7
49
2,006
darren retweeted
β‘(ΛoΜ΄ΜΆΜ·Μ€ α΄ oΜ΄ΜΆΜ·Μ€Λ)βΈ* the old Chinese Internet is almost gone, but I wanted to hold on to a piece of my childhood - so I scraped 5000 gifs from tencentβs CDN @waybackmachine to make a museum of 2005-2009 qZone (our MySpace).
you can even design your own page! link below :)
3
8
35
1,938
darren retweeted
Jun 13
the bubble is depressurising. time to focus on the skills for the next local and open phase.
your career, your startup, depend on learning these skills:
- evals
- local inference
- post-training
with this combination, you can measure how a model actually performs on a given task, and improve it needed.
if you focus a model on a use case or domain, it will be cheaper than a general API.
it's time to stop cosplaying armageddon and be practical on open models.
Jun 13
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: anthropic.com/news/fable-mytβ¦
9
9
104
8,401
darren retweeted
Jun 13
Dear Anthropic, you don't get it. The Govt is teaching you a valuable lesson about asking for regulations, and you still aren't learning it.
You don't get to say how the policies are applied, the Govt does. This should be the moment you back down from regulation... But in the post saying you think it's unfair, you double down on saying you believe the govt should be able to block AI deployments.
Stop it. Just stop it.
Jun 13
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: anthropic.com/news/fable-mytβ¦
13
18
179
14,613
15h
π
Jun 13
GLM-5.2 on KingBench (3).
Thoughts: The model has superb taste. It is greater at UX than UI. The code is always very clean. It is great at One-shot wonders. I asked it to fine-tune a whole local model and it did it in 30mins!
This is just a great model to use all-round.
1/n
2
120
darren retweeted
20h
The prospect of arbitrary and capricious government hijacking of frontier AI capabilities strengthens the case for character/virtue/principle-aligned AI and weakens the case for tool-paradigm do-whatever-they-are-told AI.
2
8
42
1,449
darren retweeted
You should all read this incredibly thoughtful meditation on generative design in real built environments, and the barriers to fully realizing its potential. The future will require more and more infrastructure, physical and digital.
May 31
Some thoughts on bringing AI into the built world and the pieces still missing
decivilize.com/anexplosionofβ¦
2
15
1,485
darren retweeted
Systems like this must exist. This is the way out.
The 8B model currently training on Agora is 350B tokens in and continuing to converge. The top level metrics and evals look almost exactly like a centralised run. But;
- 133 external contributors total bringing 4090's, 5090's, L40S/RTX 6000 and RTX 6000 Pros. These are cards that people actually own - there are no H100, B200's etc.
- The max number of nodes the system can support (104) was filled almost immediately. The authorization layer is receiving approximately 100 requests/minute to join.
- The total tokens/per second processed moves directly with amount of compute in the swarm, with Agora constantly optimising to make most efficient use of what hardware is present.
- MFU is approximately 20%, TPS is 170k tok/s. There are near constant communication failures which Agora is completely absorbing without slowdown.
- The system is effectively on auto-pilot, requiring very little intervention from us. Bad nodes are purged immediately before training is affected and new nodes take their place.
4
7
49
8,271
17h
when multi-agent actually works, it will match frontier performance and be much cheaper. closer now.
Notably, the budget panel was comparable with Claude Fable 5 in performance.
A panel of Gemini 3 Flash, Kimi K2.6, and DeepSeek V4 Pro, fused together, beat solo GPT-5.5 and solo Opus 4.8 outright.
And it landed within 1% of Fable 5 while costing roughly half the price.
2
3
384
darren retweeted
20h
mfrs out here 'i need ten biiilyun dollars to train a model!!'
rio: hold my sunscreen
>Rio 3.5 Open 397B is a frontier-class general-purpose AI model developed by IplanRIO, the municipal IT company of Rio de Janeiro's city government
huggingface.co/prefeitura-riβ¦
3
4
59
2,267
Jun 13
love is a kind of infrastructure
2
96
darren retweeted
"You're finally awake! You hit your head pretty hard there.
Huh? Gradual disempowerment? AI-assisted cyberattacks? Mythos and Fable? Listen, we just got some new 1080 Tis, let's try finetuning BERT on the GLUE benchmark!"
6
20
609
33,815
darren retweeted
15 Nov 2023
4
13
82
3,548
Jun 12
god above saw, ever in the mind
blue and white irises in a line
Jun 12
π Kimi-K2.7-Code, our latest coding model, is now released and open-sourced!
π· Improved coding & agent performance over K2.6: 21.8% on Kimi Code Bench v2, 11.0% on Program Bench, and 31.5% on MLS Bench Lite.
π· Reasoning efficiency: Less overthinking, with 30% lower reasoning-token usage compared to K2.6.
π· Long-horizon coding: Improved instruction following, higher end-to-end coding task success rates.
β‘οΈ 6x High-Speed Mode coming soon!
π Available today via Kimi API and Kimi Code.
π Kimi Code: kimi.com/code
π API: platform.moonshot.ai
5
161