Adam O'Reilly

1.1K posts

Adam O'Reilly banner
Adam O'Reilly

Adam O'Reilly

@adam__oreilly

synthetic reasoning data generation researcher

Columbus, Ohio Katılım Temmuz 2011
207 Takip Edilen149 Takipçiler
Adam O'Reilly
Adam O'Reilly@adam__oreilly·
@rowancheung I wouldn’t say unlimited cooling.. I remember Carl Schulman on Dwarkesh’s pod talking about boiling the oceans as a limit of gdp growth
English
0
0
0
174
Rowan Cheung
Rowan Cheung@rowancheung·
There's a startup trying to build data centers in the ocean. And it's INCREDIBLY fascinating: Mass consumption of electricity and water is a growing bottleneck for data centers. So by moving offshore, it eliminates both problems -- the ocean provides unlimited cooling, and the waves provide unlimited power. There are also no engines, so the data centers drive themselves to their destination by using the shape of their hull to propel through waves. Called Panthalassa.
English
104
63
606
139.7K
Adam O'Reilly retweetledi
☘︎
☘︎@vintageforestt·
june mood
English
52
8.3K
34.7K
1M
Adam O'Reilly retweetledi
Xenova
Xenova@xenovacom·
I gave Fable 5 one job: write custom WebGPU kernels for Gemma 4 inference. It climbed to 84 tok/s, then hit a wall, insisting further optimization was impossible. Hours later, Anthropic rolled back invisible LLM development safeguards, and it hit 255 tok/s. The next day, access to Fable 5 was suspended globally.
English
149
376
5.4K
1.2M
Adam O'Reilly retweetledi
Super Dario
Super Dario@inductionheads·
If we are at the point that Anthropic is degrading performance for any sort of meaningful research in biology, chemistry, or computer science… Maybe it really is over?
English
7
7
101
4.8K
Adam O'Reilly retweetledi
Sholto Douglas
Sholto Douglas@_sholtodouglas·
Programming has already changed dramatically. On my best days - I literally feel IO bandwidth limited managing the concurrent threads. As the models get better the level of abstraction will keep going up - very soon the models will only come to us for hard calls. Eventually - they won't come to us at all. LOC is obviously a terrible metric, but is correlated
Anthropic@AnthropicAI

Today, Anthropic engineers on average ship 8x as much code per quarter as they did compared to 2021-2025.

English
37
25
604
75.7K
Adam O'Reilly
Adam O'Reilly@adam__oreilly·
@blakeandersonw Thanks for sharing. Does this experience reinforce your understanding of Jesus as an awakened human or something else?
English
1
0
0
176
Blake Anderson
Blake Anderson@blakeandersonw·
I just did DMT for my second time. Will share a few of my thoughts below. Feel free to ask me anything.
English
37
1
212
43.9K
Miles Brundage
Miles Brundage@Miles_Brundage·
In retrospect this joke was too subtle for those reading quickly
English
4
0
72
3K
Miles Brundage
Miles Brundage@Miles_Brundage·
Personal update: I've decided to join Anthropic in continuing to be based in SF
English
30
6
622
38.2K
Adam O'Reilly retweetledi
Zahle Khan
Zahle Khan@zahlekhan·
@karpathy weirdest TL
Zahle Khan tweet media
English
3
37
736
132K
Adam O'Reilly retweetledi
Andrej Karpathy
Andrej Karpathy@karpathy·
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
English
8K
11.1K
150.2K
27.7M
Cursor
Cursor@cursor_ai·
Starting today, you can run cloud agents inside fully configured development environments. Set them up the same way you'd set up a laptop for an engineer: cloned repos, installed dependencies, and toolchain credentials.
English
130
166
2K
287.2K
roon
roon@tszzl·
no bro you need to turn on “/extrausage”. dawg are you sure you have “/fast” mode on? Did you check the “no mistakes” toggle? are you sure you picked “correct mode”? did you turn up the “autonomy slider”, that’s how the pros use it,
English
189
92
3.4K
201.1K
Adam O'Reilly retweetledi
fabian
fabian@fabianstelzer·
Crazy declassified 1966 footage of an Unidentified Friendly Opus (UFO)
English
501
400
4.3K
2.7M
Adam O'Reilly
Adam O'Reilly@adam__oreilly·
@nptacek As long as it passes the test suite or other objective verifier it doesn’t matter if it’s synthetic or not
English
1
0
2
355
CuddlySalmon
CuddlySalmon@nptacek·
i have never understood the synthetic data skeptics it's like they haven't played around with models at all, just stuck to whatever the prevailing view was
English
14
4
164
5.6K
Adam O'Reilly
Adam O'Reilly@adam__oreilly·
@tszzl you need to be termiusmaxxing (mine is running sweet! cli agent)
Adam O'Reilly tweet media
English
0
0
1
124
roon
roon@tszzl·
people are walking around with their laptops slightly ajar to keep their agents running
English
506
197
4.6K
749.6K
ClaudeDevs
ClaudeDevs@ClaudeDevs·
For the developers building with Claude, a direct line from the team. Follow for changelogs, API releases, community updates, and deep dives.
English
616
1.5K
21.4K
8.9M
Adam O'Reilly retweetledi
PrismML
PrismML@PrismML·
Today we’re announcing Ternary Bonsai: Top intelligence at 1.58 bits Using ternary weights {-1, 0, +1}, we built a family of models that are 9x smaller than their 16-bit counterparts while outperforming most models in their respective parameter classes on standard benchmarks. We’re open-sourcing the models under the Apache 2.0 license in three sizes: 8B (1.75 GB), 4B (0.86 GB), and 1.7B (0.37 GB).
PrismML tweet media
English
117
304
2.2K
479K
Adam O'Reilly
Adam O'Reilly@adam__oreilly·
@bstnxbt how many cli agents could you run in parallel on that machine?
English
0
0
0
37
bstn 👁️
bstn 👁️@bstnxbt·
Qwen3.5-4B just hit 133 tok/s on my M5 Max baseline MLX does 42 tok/s, so 3.2x with same output
English
3
0
63
5.9K
Adam O'Reilly retweetledi
gabrielle
gabrielle@legitimatetiger·
how it feels to accept that reality is just a carbon-based simulation where your infinite consciousness gets to play as a human avatar and remember itself as the creator of all experience despite the inherent limitations of a fragile physical existence
English
118
3.6K
26.6K
570.4K
Adam O'Reilly
Adam O'Reilly@adam__oreilly·
@a_musingcat “Ok mythos - accelerate the recognition that everything is already in perfect balance”
English
0
0
0
10