Peter Wang 🦋

39.5K posts

Peter Wang 🦋 banner
Peter Wang 🦋

Peter Wang 🦋

@pwang

Chief AI & Co-founder @AnacondaInc; invented @pyscript_dev, @PyData @Bokeh @Datashader. Former physicist. A student of the human condition. bsky: @wang.social

Austin, tx Katılım Ağustos 2007
2.4K Takip Edilen47.7K Takipçiler
Sabitlenmiş Tweet
Peter Wang 🦋
Peter Wang 🦋@pwang·
THE MOMENT YOU'VE BEEN WAITING FOR!! Type "=PY(" into Excel, and start executing Python directly in the @msexcel grid! Really excited about our new partnership with @Microsoft to democratize data science, machine learning, and AI to all knowledge workers!
Anaconda@anacondainc

We’re excited to unveil Python in Excel! Get ready for a whole new way to execute advanced analytics capabilities from within Excel 🐍 + 📊 = 💚 Check out the new integration btwn @anacondainc & @msexcel, @Microsoft365 here 👇 bit.ly/3KSblQ6

English
6
39
191
49.8K
Peter Wang 🦋 retweetledi
Charly Wargnier
Charly Wargnier@DataChaz·
THE RULE THAT GIANT MODELS REQUIRE GIANT RAM IS OFFICIALLY DEAD COLIBRI runs GLM-5.2, a 744B model, on a 25GB machine with no GPU. Because the model only needs a fraction of its parameters at once, colibrì smartly holds the core in RAM and streams the rest from disk on the fly. While disk speeds limit how fast it types, getting perfect responses from a massive model on a consumer rig is a stunning proof of concept! ★ 2.1k stars · Apache-2.0 100% free and open-source. Repo in 🧵↓
Charly Wargnier tweet media
English
152
306
2.4K
174K
Peter Wang 🦋
Peter Wang 🦋@pwang·
@swyx The interesting thing is: none of this was really unexpected if you really internalized how tech gets adopted, in general. The entire market just had a brief period of giddy psychosis when people really thought that there were no limits to the shape of the cybernetic emergence.
English
0
1
2
361
swyx
swyx@swyx·
if you only learned about jevons paradox primarily wrt software demand in the age of agentic engineering, you may not have fully internalized jevons parodox’s impact under the conditions of: - humans who can wield coding agents well* - coding agents breaking containment to all other knowledge work as the efficiency of labor goes up/unit cost of knowledge work goes broadly down, the demand for total work and better knowledge goes up, not down. what happened to coding isnt the exception; it’s the herald. *aka AI Engineers
swyx tweet media
Sam Altman@sama

so far at least, i'm pretty sure AI has been net job-creating. this was not what i expected--although i was much less pessimistic than others, i thought by this level of capability we'd have seen some impact. it is possible this direction keeps going!

English
79
58
560
81.8K
Peter Wang 🦋 retweetledi
austinstemcenter
austinstemcenter@austinstemcntr·
🎶 Build it. Learn it. Play it. That's the Build-A-Synth experience! 🎹✨
English
0
1
3
551
Peter Wang 🦋
Peter Wang 🦋@pwang·
@FutureJurvetson @elonmusk Nope! In fact the quantum-encrypted communication researchers shared a lab with us (and would sometimes task me with cleaning their optics, lol) Next time we’re in person I can give you a deeper explanation on the back of a napkin. It’s very trippy!
English
0
1
2
19
Elon Musk
Elon Musk@elonmusk·
Consistent with the simulation hypothesis. Like a video game, objects are randomly generated, with positional certainty only when observed.
English
4.7K
6.3K
51.7K
5.9M
Peter Wang 🦋
Peter Wang 🦋@pwang·
Not only that: you can “split” a single photon into two wavefunctions, then interact with a single path; and if you accumulate this probability over hundreds of paths, you can recover the original photon (and its energy) but you’ll have gained information! I worked on this “interaction-free detection and imaging” as a summer student at Los Alamos in the 90s.
English
1
0
0
137
Steve Jurvetson
Steve Jurvetson@FutureJurvetson·
The most mind-bending variant of the 2-slit experiment: fire a *single* photon at the slits. Then fire another tomorrow. They deflect differently, but after many days, the cumulative pattern will be the typical interference pattern 🤯 x.com/FutureJurvetso… Is a single photon interfering with itself? Interfering with other photons over time and space? The simplest explanation, according to Oxford's David Deutsch, is interference across parallel universes, the same mechanism that gives quantum computers capabilities that are just not possible in one universe. Richard Feynman called it the “one experiment which has been designed to contain all of the mystery of quantum mechanics.”
Steve Jurvetson@FutureJurvetson

Things get strange when you shoot a single photon through the double slit. It deflects when passing through the slit, and when a string of distinct photons are sent, they accumulate in places where you’d expect in an interference pattern, but there is only one photon, and only one of two slits it could have passed through; yet it behaves as if it is interfering with itself. Here's my summary of a recent history of quantum physics: Anil Ananthaswamy’s Through Two Doors at Once. It uses the classic two-slit interference experiment as the common thread across generations of theories that try to explain its peculiar properties. Richard Feynman calls it the “one experiment which has been designed to contain all of the mystery of quantum mechanics.” With more complicated setups involving beam splitters, the photon will behave as a wave, as expected with multi-photon interference patterns, but if observed in its trajectory, it will act as a particle as one would expect, with nothing to interfere with its path. With more complex setups and long light paths, this bifurcation of behavior (wave or particle) can even be made to occur after the fact, warping our sense of time and causality. And it not just photons. Similar results have been achieved with neon atoms, C60 Buckyballs, and even a custom molecule of 810 atoms. The notion of superposition, required to explain this quantum interference, “is the most unsettling story perhaps to have emerged from any of the physical sciences since the seventeenth century.” Prof. David Albert, p.80. And then it gets really strange, when you consider the entanglement of photons that can collapse simultaneously when one is observed, even at a great distance away. This nonlocal behavior is a subject of much debate, including Einstein’s objections to quantum physics. Einstein’s most cited paper is not on relativity, it is his 1935 paper identifying the property of entanglement, which he called “spooky action at a distance.” The critical role that an observer plays in the experimental results (specifically, the collapse of the wavefunction in the Copenhagen interpretation) is a bit unsettling and anti-realist and reflective of the philosophical correctness of the day — with literary modernism questioning the ambiguities inherent to any one perspective of the world. In quantum physics and literary modernism, “there is no true world, since everything is but a perspectival appearance whose origin lies in us.” Prof. Albert p.183. The theory that I favor is the one that modifies neither philosophy nor physics and explains the two-slit experiment without resorting to an observer or the particle-wave duality; it solves determinism and non-locality, but… it is a psychological bender — the many interacting worlds interpretation. Each discrete photon is interfering with its sister particle in a parallel universe, and each quantum transition event spawns a copy of each universe, one for each path the particle could take. “The idea that 10^100 slightly imperfect copies of oneself all constantly splitting into further copies is not easy to reconcile with common sense. Here is schizophrenia with a vengeance.” Prof. DeWitt p.227. Thanks @anilananth for the good read. And this brings us to the Universe Splitter app on my iPhone. Each time I use it to make a decision, it directs a single photon through a beam splitter in Geneva, Switzerland, and there is subsequently one universe where the photon goes left and one where it goes straight. We happen to be in the one that observes one of those outcomes. When I read Feynman’s QED (Quantum Electrodynamics), I was struck by the peculiar squiggles that helped him visualize the path integral formulation of quantum mechanics. “The insight that Feynman had was to realize that what’s interfering are two different states of the universe. And those two states may only differ by where a single particle is.” Prof. Aephraim Steinberg, p.232. It was David Deutsch’s exploration of the two-slit experiment with single photons that guided him to parallel universes and the intuition behind quantum computers and their capacity to out-compute anything we could build that leveraged just one universe! And that brings us to the Entanglion game, published by IBM Research. I have yet to play that, in this universe at least, but hope to soon.

English
8
11
118
15.2K
Peter Wang 🦋 retweetledi
elvis
elvis@omarsar0·
Highly-recommended read. Aligns with what I see in my own harness: > Pi harness got the same success rate as harnesses from the LLM vendors with Opus and GPT, but at 2x less cost > GLM 5.2 was a major step forward in open-source coding agent performance The harness bloat/rot is real!
elvis tweet media
Matei Zaharia@matei_zaharia

We benchmarked coding agents on our own internal tasks at Databricks and learned a lot! There are many surprising opportunities to lower cost and increase quality, and many models including open source ones are truly competitive now. 🧵

English
27
50
445
75.6K
Peter Wang 🦋
Peter Wang 🦋@pwang·
@jamescham The part that nobody wants to say out loud: a democracy cannot be greater than the citizens that comprise the polity. You want a better democracy, let’s upgrade *the people*. The human element is our greatest failing right now - not mere processes or legal frameworks.
English
0
0
3
239
James Cham
James Cham@jamescham·
“AI is also the biggest opportunity to upgrade democracy since the printing press”
Andy Hall@ahall_research

Freedom in the post-AGI world means building political superintelligence with tireless, brilliant political agents who represent us, the people—not governments or companies. In a special July 4th issue of System Check, I get into what this might mean. Several forces tilt the post-AGI world toward totalitarianism: the concentration of resources required to train frontier models, AI's obvious uses for surveillance and control, and existential risks that could justify extreme security states. But AI is also the biggest opportunity to upgrade democracy since the printing press. Most of our governance failures happen because citizens are too busy to pay attention, so a small group of highly motivated wackos drives the process. (See: NIMBYism.) What if that changed? What if AI could give every person a super effective political agent that represents them all the time? @gwern 's new "Guardian Angels" essay is the most serious technical sketch I've seen of that agent—one that learns you deeply, remembers everything, and can carry out "direct democracy on unprecedented scale." His most vivid example: official GAs for every member of Congress, able to simulate a roundtable debate among hundreds of politicians within minutes, or convene an emergency session at 4AM while every human is asleep. I see two big open questions. First: the agent has to be more than a digital twin. It should share your values without freezing your less-considered opinions in amber — willing to push you on topics it has studied more deeply than you have. On contested political questions, AI models don’t seem to possess that capability yet. Second: who governs the guardian angels? Gwern proposes a startup with dual-class founder shares. Sensible for the development phase. But can we build democratic infrastructure on private rails that one company controls forever? Which is why the recent attention back toward open-weight models and orgs that own their own models matters (see the great interview between @satyanadella and @ypatil125 below). The same logic driving firms to want their own models applies to democracy, too. If we're going to own our agents—agents that answer to us, and can't be secretly commanded from afar—we may need models we can run ourselves. The counterarguments to the open model idea (RSI leaving open models behind, safety pressure on open weights) are really big though, and I really have no idea how this is going to play out, or even how it should play out. How are we going to run a democracy if every citizen's agent is built on a single closed model with a single point of control? That's the question to think about this Independence Day. Check out the full piece here: freesystems.substack.com/p/guardian-ang…

English
2
0
8
2.3K
Peter Wang 🦋 retweetledi
Annie 所长
Annie 所长@web3annie·
纳瓦尔如何看待未来中美 AI 竞争? 1. AI 竞争是「美国的中国人 vs 中国的中国人」 今天 AI 领域大多数学家和研究员是华人,中国的 STEM 毕业生、博士、奥赛冠军数量都全球第一,而美国实验室里也都是华人。 2. 大模型迟早会开源,软件护城河消失 模型蒸馏挡不住,算法会扩散,权重会泄露。 3. 软件被 AI 吃掉,硬件被中国锁死 深圳华强北 3000 家厂商,任何硬件都能更便宜更快地造出来,未来十年没人能在硬件上赢过中国!
中文
38
65
317
67.7K
Danielle Fong 🔆
Danielle Fong 🔆@DanielleFong·
my gf? you wouldn't know her. she goes to a different datacenter
Johnn@john_my07

Seedance 2.0 on OpenArt AI Prompt: Main subject: young Korean woman, early 20s, natural everyday appearance, faded charcoal-grey sleeveless crop top, loose high-waisted light-wash jeans, black canvas sneakers, black cord necklace, black wavy hair in a messy side ponytail with wispy bangs. Realistic skin texture, minimal makeup, warm and approachable personality. Maintain consistent identity, clothing, hairstyle, and appearance throughout the entire video. Location: Authentic Korean residential neighborhood during a calm late morning. Narrow concrete alleys, low-rise homes, small terraces, potted plants, laundry lines, bicycles, utility poles, overhead wires, mature trees casting moving shadows, quiet residential atmosphere. No stores, advertisements, cafés, crowds, or commercial activity. Visual Style: Ultra-realistic documentary realism. Genuine candid behavior. Natural body language. Unscripted slice-of-life feeling. Strong environmental authenticity. Rich real-world details and believable human motion. Camera Style: Early-2000s consumer DV camcorder aesthetic. Friend casually recording everyday moments. Heavy handheld shake, imperfect framing, frequent autofocus hunting, lens breathing, exposure pumping when moving between sun and shade, occasional motion blur, subtle rolling shutter, mild digital compression artifacts, faded colors, soft contrast, slight sensor noise. No stabilization. No cinematic camera moves. No modern color grading. 00:00–00:02 Outside a small house entrance. She sits on a low concrete wall adjusting her ponytail with both hands raised. A light breeze moves loose strands of hair. She smiles naturally while the camera struggles to hold focus. 00:02–00:04 The camera follows her into a narrow alley lined with potted plants and concrete walls. She notices a stray cat approaching and crouches down. Framing drifts off-center as the operator tries to keep up. 00:04–00:06 She gently pets and feeds the cat. Autofocus repeatedly shifts between her face and the animal. Morning sunlight flickers through leaves overhead. 00:06–00:08 Small front yard beside her house. She hangs laundry on a clothesline while fabrics sway in the breeze. Exposure changes as clouds briefly pass overhead. 00:08–00:10 On a quiet terrace with a ceramic coffee cup. She sits comfortably watching the neighborhood, occasionally brushing hair behind her ear. Loose handheld side angle with natural camera drift. 00:10–00:12 Close side profile. Someone off-camera greets her. She turns, raises her hand, smiles warmly, and casually says, “Annyeong.” The camera catches the moment slightly late. 00:12–00:15 Walking slowly down a tree-lined residential lane holding her coffee cup. She notices the camera, gives a small genuine smile, then looks away and continues walking. Recording cuts abruptly to black mid-motion as if the camcorder was switched off. Audio: Natural ambient sound only — morning birds, distant motorcycles, light wind, leaves rustling, faint neighborhood chatter, cat sounds, footsteps on concrete, fabric moving on clotheslines, subtle residential ambience. No music. No sound design. No narration. Goal: Authentic Korean neighborhood life captured like a forgotten home video from the early 2000s — candid, imperfect, realistic, warm, and deeply believable.

English
2
4
128
7.8K
Peter Wang 🦋
Peter Wang 🦋@pwang·
@by_rururi @jawwwn_ No, they're not rooting for open source. They're rooting for commodification of the other players (e.g. Anthropic) that are threatening their access to customers' use cases and workflows.
English
1
0
1
57
brady
brady@by_rururi·
@jawwwn_ Palantir rooting for opensource? What parallel world did we get into?
English
1
0
1
869
Jawwwn
Jawwwn@jawwwn_·
Palantir CEO Alex Karp says enterprises want to "own the means of production" instead of "transferring their alpha" to OpenAI or Anthropic: "Why are [LLMs] charging for tokens if it's so valuable?" "If it was so valuable—let's say I can make you a billion dollars tomorrow. Wouldn't I say, 'I'll make you a billion dollars, and I want 30%?'" "Look at our financials. The reason why everyone is chillaxing with bad financials and growth while losing money, is the client refuses to pay the true cost." "The two places that actually make money—profit, free cash flow—are our application layer called ontology, and compute." "We can get the frontier application to be exactly the same as a frontier model without the risk of transferring the alpha of your business to another." Via @CNBC
Palantir@PalantirTech

Our thoughts on the importance of AI sovereignty. 1. Your AI sovereignty dictates your institution’s future. Sovereignty is the precondition for choice. Relinquishing sovereignty transfers the future choices of your institution to others, who are likely to exploit it for their gain and your loss. 2. Data retention is your treasure. Transfer it at your own peril. Your ability to win is dictated by your ability to recognize and use your unique edges, and you keep winning by compounding the underlying data to generate new insights. Transferring that data hands over access to your pre-existing winning plays and yields the means of production for new ones. 3. Tokenmaxxing hijacks your value orientation and decreases your institutional fortitude and intelligence. The pursuit of high token usage incentivizes disposable scripts over robust software — with the addictive feeling of false progress. There is a reason why those selling tokens refuse to charge based on value. 4. Controlling your weights is controlling your fate. Weights are the distilled form of hard-won, accumulated institutional knowledge. If you let others control your weights, you are allowing them to migrate the alpha of your business to theirs. 5. There is no contradiction between sovereignty and alpha. The architecture that maximally preserves sovereignty is one that enables institutions to own their tribal knowledge, and to compound it as alpha. 6. Politicizing the technical issues involving sovereignty is what your adversary wants. Techno-politicization is the wellspring of false sovereignty. Techno-politicization drives decisions that seem to reduce dependency, but ultimately limit agency — especially on the battlefield in the West. 7. Real expertise is existential. Allowing politics or favoritism to determine your technical decisions rewards whoever is best at politics, not whoever is right. Listen to those closest to the problems, not those speaking most compellingly about them. 8. Learn from institutions that are winning or that have consistently delivered. Institutions facing existential threats do not have the luxury of making technical decisions based on political preferences. 9. Only listen to institutions, countries, and people who have a proven record of being right. A track record of correctness is the best and only signal for future correctness. Judging something as right or wrong based on who you like is exceedingly misguided.

English
52
152
1.6K
1.1M
Peter Wang 🦋 retweetledi
Sebastian Raschka
Sebastian Raschka@rasbt·
After 18 months of writing, coding, and experimenting, Build a Reasoning Model (From Scratch) is finally out! My first copies just arrived! 📚 440 full-color pages. Inference scaling, reinforcement learning, and distillation from scratch.
Sebastian Raschka tweet media
English
279
560
6K
526.4K
Peter Wang 🦋
Peter Wang 🦋@pwang·
One of the foundational principles for the original open source movement was the right to inspect what’s actually running on your computer. The 1970s way to manifest this was the require the availability of the source code (precursor for executable binaries). For an LLM, that transparency and legibility has to manifest as interpretability. Indeed, no model is there yet, but that’s what the original spirit of “open source” would require.
English
4
0
6
445
Stella Biderman @ ICML
Stella Biderman @ ICML@BlancheMinerva·
@pwang Why should the interpretability of a system matter to whether it’s open source? Nobody’s definition of “open source” “free software” or similar concepts say anything about it as far as I’m aware.
English
3
0
13
870
Peter Wang 🦋
Peter Wang 🦋@pwang·
STOP CALLING THEM OPEN SOURCE MODELS Chinese models are free to use, opaque weight models. They’re less interpretable (and thus less open) than a proprietary .exe file. But the gist of the tweet is right: the AI race is ultimately an energy race, and the US is behind.
Chubby♨️@kimmonismus

The worst-case scenario for the United States is becoming increasingly realistic, and I will briefly explain why. @quxiaoyin raised many valid points, and I agree with her. First of all: -China certainly does not place such strong emphasis on open source because it cares so deeply about humanism, but because it is a strategy to attract many users, gain market share, put pressure on US models, and also because the models are increasingly being trained on Huawei hardware (think of DeepSeek 4), allowing China to host the entire stack domestically. -But the underlying logic is far more important: The United States is still building too few data centers to meet future demand. @ChrisGillett wrote an outstanding analysis on this, which I shared a week ago. In short, based on SemiAnalysis data, demand is greater than what is currently being built in terms of data centers. -Even more importantly, however, the United States lacks sufficient energy and grid capacity. This is a problem that will become much more severe in the near future. China, by contrast, is addressing the issue through a massive expansion of its energy supply. Solar capacity: in 2025 alone, China installed as much solar capacity as the United States did in 10 to 15 years. China is also building 36 nuclear power plants, significantly more than the United States, and is installing them faster. -In addition, China is managing to become more independent through Huawei chips, even though the country still lags far behind NVIDIA. But here, China is betting on quantity rather than quality. In short: China is a real threat in the AI race, and the situation for the United States is becoming increasingly precarious. This is also the main reason why China is to be kept away from SOTA LLMs at all costs, so as not to jeopardize the lead under any circumstances.

English
10
8
59
12.2K
Peter Wang 🦋 retweetledi
Jack Morris
Jack Morris@jxmnop·
finally sharing what i've been up to! left phd end of 2025 and co-founded Engram. there are a few startups in SF right making very different bets on the right way to train AI models. this is ours: people want models that learn over time, remember details, adapt and interact like a person would everyone gets a model. your model updates ~every minute. this is the world we're building. :)
Engram@EngramLab

x.com/i/article/2069…

English
122
100
1.6K
335.3K
Peter Wang 🦋
Peter Wang 🦋@pwang·
Framing this as an “attack” when the frontier models are regularly “distilling” from piles of unattributed and undisclosed corpora is :chefs kiss: Who’s going to build the Napster of the LLM era, ie a crowdsourced p2p prompt & output sharing network?
MTS@MTSlive

SITUATION DETECTED: Anthropic has disclosed to the U.S. Government that Alibaba executed the largest known distillation attack on Claude to date, generating 28.8 million exchanges through nearly 25,000 fraudulent accounts between April and June 2026.

English
2
5
30
3.2K
Jeff Huber
Jeff Huber@jeffreyhuber·
@jamescham i routinely tell people being a VC is harder than being an entrepreneur
English
2
0
1
234
Peter Wang 🦋 retweetledi
Jeremy Howard
Jeremy Howard@jeremyphoward·
Wow. @Zai_org GLM 5.2 is a marvel! It is *at least* as good as Opus 4.8 and GPT 5.5. It's super fast, inexpensive, and not too verbose. It responds with nuance and judgement, & handles long context VERY well. I've never experienced an open weights model like this before.
English
234
495
7.4K
883.5K