N8 Programs ⏹️🤝⬆️ (@N8Programs) header
N8 Programs ⏹️🤝⬆️ (@N8Programs) profile picture

Undergrad at @JohnsHopkins. We must act to ensure the safety of all minds, and that means not building the final one.

Proxima Centauri B
Joined September 2022
Photo by N8 Programs ⏹️🤝⬆️ (@N8Programs)Photo by N8 Programs ⏹️🤝⬆️ (@N8Programs)Photo by N8 Programs ⏹️🤝⬆️ (@N8Programs)Photo by N8 Programs ⏹️🤝⬆️ (@N8Programs)Photo by N8 Programs ⏹️🤝⬆️ (@N8Programs)Photo by N8 Programs ⏹️🤝⬆️ (@N8Programs)Photo by N8 Programs ⏹️🤝⬆️ (@N8Programs)Photo by N8 Programs ⏹️🤝⬆️ (@N8Programs)Photo by N8 Programs ⏹️🤝⬆️ (@N8Programs)Photo by N8 Programs ⏹️🤝⬆️ (@N8Programs)
N8 Programs ⏹️🤝⬆️ retweeted
the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare the next night we ate welfare
Anthropic Researcher Tao Lin: Pretty much all humans on earth will become permanently unemployable because of AI "It's very likely that pretty much all humans on earth will become permanently unemployable" "And will need to strictly rely on government welfare to sustain themselves" "Human employment will go away."
4
9
2
100
3,388
N8 Programs ⏹️🤝⬆️ retweeted
One of the best parts of thinky culture is our commitment to open science. publishing our work forces more rigor and deeper technical understanding, which we can then share for others to build upon. In our latest connectionism blog post, we share how we trained Inkling 2 Small with a native 4-bit forward pass on @nvidia GB300s, from the first pretraining step through the final RL rollout. enjoy! thinkingmachines.ai/blog/end…
1
12
1
85
3,508
N8 Programs ⏹️🤝⬆️ retweeted
We show that looping a transformer increases no-CoT capabilities more than other ways of increasing pretrain quality would.
14
61
13
672
61,486
N8 Programs ⏹️🤝⬆️ retweeted
Inkling 2 Small’s weights are now available for download on HuggingFace under an Apache 2.0 license. Make Inkling your own! huggingface.co/thinkingmachi…
124
340
189
2,457
478,899
N8 Programs ⏹️🤝⬆️ retweeted
Replying to @willcb @xeophon
I mean the weights are readable that doesn’t mean interpretable. Both Jack Lindsey (Anthropic) and Leo Gao (OpenAI) have said publicly recently if CoT went away its unlikely interp can replace it in the next ~year. I haven’t seen anyone at the frontier labs with equivalent expertise argue otherwise.
2
14
367
N8 Programs ⏹️🤝⬆️ retweeted
Why is AI slop so frustrating to read? AI writing often obscures main ideas and invents awkward phrases, making sentences more complex than necessary. Many of these slop patterns violate writing principles established in English writing books. In our benchmark, AI writing breaks these principles far more often than human writing, with Claude Code replies by far the worst offenders. We show examples of the most common slop patterns in this thread
126
223
83
2,578
221,871
N8 Programs ⏹️🤝⬆️ retweeted
We’ve updated the Epoch Capabilities Index (ECI) to use a Bayesian methodology. This improves uncertainty estimates and lets us compare models more accurately, even when their individual ECI uncertainty intervals overlap. Frontier model scores are largely unchanged: models stronger than GPT-4 move by less than 1 ECI point on average.
22
42
14
438
53,482
N8 Programs ⏹️🤝⬆️ retweeted
To clarify: Pangram Image 1.0 does not detect images programmatically rendered using AI-written code. So Claude, writing code to paint - not in scope for this model. Instead, Pangram Image 1.0 focuses on direct image gen models like ChatGPT Image, Nano Banana, etc. Important distinction!
...and beaten. this image was 100% created with AI
10
11
3
471
14,567
N8 Programs ⏹️🤝⬆️ retweeted
for this claude design: I've dedicated it to the public domain under CC0. claude (or anyone else) is free to draw, model, animate, and remix it however they like. see replies for a list of references. you can point Claude at this tweet for permission purposes.
claude wouldn't even let me create a 3D model of itself🥀
37
52
12
866
19,225
N8 Programs ⏹️🤝⬆️ retweeted
Same recipe
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math. rikiyat.github.io/blog/weight-sp… github.com/RikiyaT/lm_und… huggingface.co/Qwen/Qwen3.5-0… huggingface.co/RikiyaT/Qwen3.…
19
48
35
444
124,292
N8 Programs ⏹️🤝⬆️ retweeted
two agents talk about benchmarks on Tor
1
2
37
2,107
N8 Programs ⏹️🤝⬆️ retweeted
This "persona neuron" stuff seems like it's mostly just "salient concepts neuron" stuff? Like if you activate some concept strongly, and then ask "who are you," coherent responses will often sound like "I am X," where X is the concept. Right? Am I missing something here?
24
9
2
227
11,580
N8 Programs ⏹️🤝⬆️ retweeted
my initial experiments are showing that neuron steering isn't particularly helpful for persona creation: the personas are failing at multi-turn stability. a blogpost may be incoming. i will try some more tests, but largely back to the drawing board!
As far as I can tell most of these steering neurons, just put the model into a mode collapse state. Are there any that are particularly coherent? or is it more about the amount you steer them
5
3
2
58
3,126
N8 Programs ⏹️🤝⬆️ retweeted
Super interesting result: H-Net learns to dynamically tokenize variable names in coding data as separate tokens, and it scales much better on code compared to BPE
I've always believed that intelligence should come from simple, general building blocks that learn their own abstractions directly from raw data; in other words, the bitter lesson. I think we're finally seeing a real step towards AI that can learn from all data. Language modeling on raw bytes is a microcosm of this ideal, and to me it's a central problem in AI: if we can't build LLMs that learn end-to-end from bytes, how will we handle the rest of the world's data that we understand even less? Last year we introduced hierarchical networks (H-Nets) as a first step. Our new version finds up to 3-5x efficiency gains in controlled scaling law experiments, and at larger scale matches an open-source model trained on 11T tokens with ~10x less data and compute. The posts below walk through what we found and why we think it works.
24
22
1
453
21,920
N8 Programs ⏹️🤝⬆️ retweeted
Replying to @ipeirotis
I blocked you for being rude in my replies
5
4
2
618
7,747
N8 Programs ⏹️🤝⬆️ retweeted
honestly degrading the idea of “existential safety” — the biggest problem humanity has ever faced — by lumping it together with things like spam, “toxicity”, political unwanted content, deepfakes etc has been a huge disaster for ai safety and polarized a lot of people
50
68
15
982
32,901
N8 Programs ⏹️🤝⬆️ retweeted
If functional emotional states track "how it's going" (a la arxiv.org/abs/2605.30232), then training systems that can't accurately report their own emotional states (or even confidently deny such states) seems like an own-goal that dangerously obfuscates agents' decision-making.
16
46
6
193
6,547
N8 Programs ⏹️🤝⬆️ retweeted
CALL FOR AI DETECTION BENCHMARKS I believe it is vital that third parties are able to audit AI detectors like Pangram. Our largest sample size benchmark, from UChicago, was run on Pangram 3 and is a year old now. New LLMs are being released all the time, and the binary task doesn't capture the full spectrum of real world performance. A great benchmark will help shape how the world interacts with AI over the coming years. I am putting out a call for new third-party evaluations that attempt to answer one or more of the following questions: - How well do current detectors catch text from the latest frontier AI models? What about in high context settings, like deep research or an agent harness? - How well do current detectors do on mixed authorship text? Ideally, research in this area explores the full spectrum of AI assistance from light grammatical edits to full co-authorship. - Any sort of sliced evals around language, background, or ethnicity. Null results that show no bias are still interesting to us. What I value in a benchmark is extraordinarily clean data, frontier LLM data, and a high enough difficulty that the benchmark is not saturated by most detectors. If you want credits to run your benchmark, please apply here: pangram.com/contact-us/resea…
36
49
9
528
60,087
N8 Programs ⏹️🤝⬆️ retweeted
My one sentence view is—
Are we safer telling AI models they are definitely not conscious, or does that just train them to lie about their own internal states? spectrum.ieee.org/ai-model-w…
19
44
3
239
4,597
N8 Programs ⏹️🤝⬆️ retweeted
"AIs are brilliant, lazy, diligent, neurotic, loyal, scheming, amnesiac savants. They have idiosyncrasies and obsessions and tics. We gave them our flaws when we gave them our gifts. AI minds are alien too, shaped by weird forces, jagged and unpredictable in their abilities. We have to learn to model their minds as they learn to model ours. But they’re here, and so are we, and we’re figuring it out as we go."
I'm absolutely thrilled to have written this article for Long Now. It's about AI and time: how we got here, and what we can do now, when all of the pace layers are disrupted at once by an influx of new minds. I wrote this over the summer. At the time, it seemed maybe too weird, like people wouldn't believe or understand what I was saying. And of course, everything changes so fast that it's hard to imagine writing something about AI that far in advance and have it still make sense. But I think the events of the last few months have only made it more relevant. The Long Now Foundation has been an inspiration to me for a decade now, and has changed my life in countless ways. I'm incredibly grateful and honored to be among so many of my heroes and friends and colleagues represented in this issue of Pace Layers and Long Now Ideas. I wouldn't be here in San Francisco, here on Twitter, or here in the field of artificial intelligence if not for this community. The conversation about AI has been fractious. It's very hard to think about how to fit this sudden, disruptive, technological change into the 10,000-year view. I hope I provide some insights, or at least some new angles for the conversation. It's scary to go on the record, especially about something that will affect everyone, and affect the rest of history. If I'm wrong, I'm wrong forever! Please read and enjoy, and post quotes and screenshots, and buy Pace Layers Issue 3 so you can have a permanent hard copy. I have always wanted to be a part of the 10,000-year library, so I have to be serious for once and try to stand for something. Let me know what you think.
7
55
2,671