←
Simon J.K. Pedersen8,540 posts
Simon J.K. Pedersen avatar

Simon J.K. Pedersen

@simped

SharePoint Dev/Architect at day, Azure, F# and Node.js hacker at night.

◷ Joined April 2009▧ 71 Photos and videos
Simon J.K. Pedersen avatar
Simon J.K. Pedersen@simped·Nov 16

This could check out

The escalation of the Megaspeed crisis was never meant to reach this level. Initially, the strategy from Washington’s China hawks was remarkably straightforward. Guided by a formal letter from Republican Senator Tom Cotton, they simply wanted the Commerce Department to investigate and blacklist the front company. Had Jansen Huang accepted the loss, severed ties with Megaspeed, and ended his association with Miss Alice Huang, the entire controversy would have quietly faded a year ago. Instead, the crisis metastasized. As nonproliferation advocates attempted to close the regulatory loopholes through official channels, they collided with two insurmountable barriers. The first barrier was the corporate fortress of Nvidia. For ten months beginning in late 2025, Huang weathered intense internal compliance audits, consistently denying any romantic involvement with Miss Huang. He took his defense directly to the public, appearing on television to personally guarantee Megaspeed’s integrity. He insisted that rigorous on-site spot checks had been conducted and that all 86,000 GPUs were strictly accounted for. Behind the scenes, the pushback was even more aggressive, deeply frustrating Washington investigators. Nvidia's General Counsel, Tim Teter, actively intervened in the Bureau of Industry and Security (BIS) drafting process, inserting his own revisions and demanding access to sensitive government files to disrupt the case. This corporate stonewalling found an ally in Jeffrey Kessler, the Commerce Department's Undersecretary for BIS, who deliberately stalled and ultimately halted the effort to blacklist Megaspeed. The second barrier was President Trump. With the Commerce Department paralyzed, intelligence agencies stepped in, utilizing surveillance to confirm the secret romance between the two Huangs. In the spring of 2026, Secretary of State Marco Rubio took decisive action, overseeing a highly classified memo that identified Miss Alice Huang under the moniker of a “Workers’ Stadium female soccer fan.” The pivotal moment came in May. As Air Force One touched down in Anchorage for a refueling stop en route to Beijing, Rubio personally delivered the intelligence memo to President Trump. He explicitly warned the President against allowing Old Huang to join the delegation. Yet, after reviewing the classified briefing, President Trump disregarded the advice and permitted Old Huang to board the flight to China anyway. Now, as late November approached.time had officially run out for the hawks. The urgency, as highlighted by a Bloomberg White House correspondent, comes from a critical impending deadline: next week, President Trump departs for his Asia tour, marking his third in-person summit with the Chinese president in 2026. For the China hawks, the nightmare scenario is imminent. Their internal warnings have been suppressed, and official investigations are frozen. They feared that as the President Trump sits down at the negotiating table with Xi Jinping in Shenzen, Jansen Huang would have successfully lobbied him to use advanced chip exports potentially even the cutting-edge Blackwell architecture as a diplomatic bargaining chip to secure a summit deal. If that happens, the United States' strategic firewall on global computing power will instantly collapse. Out of time and locked out of every conventional administrative channel, the Chinese hawks lead by Rubio faction and backed by Israeli lobby did their final maneuver: leaking the Rubio spiced up Memo on Jansen to Democratic-leaning mainstream media outlets to force a public scandal, flip the political table, and purge the President's inner circle of compromised advisors.

Kevin Bass avatar
⟳ Simon J.K. Pedersen retweeted
Kevin Bass@kevinnbass·Nov 15

I've been combing forensically through the media reporting on the Jensen Huang debacle, and I'm shocked. - New York Times directly claimed at least 11 times that Alice Huang was a Chinese spy. - The Times is walking this claim back after U.S. intelligence would not corroborate it. - If real, whoever handed over the alleged intelligence to the Times potentially violated the Espionage Act, unless authorized to do so. - If Trump authorized it, why did he? - If Trump did not authorize it, why did the official violate the Espionage Act -- potentially prosecutable as a felony? - Who was the official who leaked the classified intelligence? - Was it Rubio, who authored the memo on the original CIA brief, or was it someone else? - Jensen Huang's apology for being unfaithful does not name Alice Huang, or anyone else. - The media's characterization of Jensen Huang's relationship with Alice Huang as being romantic is also based on anonymous intelligence, but is not established as fact. - Wall Street Journal fabricated a claim that Singapore investigated Alice Huang for espionage, when Singapore actually investigated her for false claims; WSJ still has not retracted its false reporting. Thus: - Jensen Huang may have had an extramarital affair with Alice Huang, or he may not have. - Alice Huang may be a Chinese intelligence asset, or she may not have been. - According to third-hand accounts of the intelligence briefs, the U.S. government suspected that Alice Huang may be a Chinese asset for well more than a year. - Yet, according to Trump's account, Jensen Huang had never been informed. - Jensen Huang, therefore, is apparently being held accountable for an extramarital affair that he may not have had, with a woman who may not have been an intelligence asset, who he had never been told was or was not such an asset. - This is nonsensical. How, after all, can he be held accountable for an affair with a Chinese intelligence asset (who might not actually be a Chinese intelligence asset) that the U.S. government knew about -- but did not warn him about? - Is not the accountable party the U.S. government official who failed to warn him? Or is he being held accountable for that affair? - Remember, Jensen Huang acknowledged that he had been unfaithful, but not with whom. - Is he being held accountable for another affair? - Unnamed sources also mention an affair with a direct subordinate Kimberly Powell. - This affair is again unconfirmed. - A romantic relationship with a subordinate violates company policy. - But the board actually has discretion about how to handle this. They could slap him on the wrist, impose some disciplinary regime, therapy, etc. - But they didn't. So to say that they were obligated to remove him is not correct. - We know that some members of the board were helicoptered in from Wall Street just 6 months before the ouster. - What role did they play? - Was the takeover by the board being planned far in advance of the reporting? So they chose to remove him. Or did they? - Jensen Huang resigned. - This is not the same as the resignation being voluntary. - What are the circumstances of his resignation? - Was he forced to resign, or did he do so from exhaustion, or to protect his company? - Furthermore, what was the flash point? Was it for the extramarital affair that may not have been one, with the Chinese intelligence asset, who may have not been an asset, that U.S. intelligence knew about but did not warn him about? Or was it about the affair with the subordinate that the board had discretion over? - We do not know the answers to any of these questions. - Unnamed sources, reported by Reuters, implicated a second Nvidia employee extramarital, also unconfirmed. - Which affairs are real? Which are mere hearsay? - How did this play into deliberations? - The media continues unabated character attacks on Jensen Huang. - Yet we also know that, for all the dirt that they try to dig up, most of it has remarkably little substance. - Alice Huang may have smuggled some state-of-the-art (banned) chips into China. - What did Jensen Huang know about this? We don't know that yet. - If these chips were smuggled, objectively, these chips are still a drop in the bucket compared to the overwhelming compute advantage that the United States has over China. Overall, the story is actually extremely thin, and we know remarkably less than the media implies. And, in my opinion, that is the story itself: a full-on media assault on Jensen Huang in the court of public opinion, on the basis of remarkably little substance. This reveals more about the media than it does about Jensen Huang. They want him gone. But why? And how does this relate to the motives of the U.S. government official who leaked the Rubio memo about Alice Huang? We also know: One of the first reporters involved at the FT says that she was tipped off to this story by someone in the tech industry. We have many more questions than answers. Personally, and my opinion at this point is clearly partisan, but I smell a room full of rats.

Simon J.K. Pedersen avatar
Simon J.K. Pedersen@simped·Nov 13

It's Friday, and my email is working just fine - yes building a centralized internet was maybe not the smartest move.

Simon J.K. Pedersen avatar
Simon J.K. Pedersen@simped·Nov 9

What a bomb

His actual 168-word statement just went live on Nvidia's official newsroom page (nvidianews․nvidia․com). Zero PR spin. He admits the affair in the very second sentence: "I owe my family and all of you an apology. I was unfaithful to my wife. My actions hurt her and our children. There is no excuse for what I did. I am deeply sorry."

Post attachment
Simon J.K. Pedersen avatar
Simon J.K. Pedersen@simped·Nov 6

Love this take. Just because we have ai it's not an excuse to get lazy. Focus on the details, now you have the time for it, because ai can do the mundane work. Everyone can build lazy shit with AI, but very few can build an maintain quality, and those who can, will win.

一个人用 Claude Opus 5.5 搓了个网页版战斗机模拟器,直接被 Three.js 官网挂在了首页最顶上。 两周 2.5 万人玩,每天还有 1000 人在线开飞机。 看完这个哥们的复盘,我觉得现在用 AI 做独立产品,真正的分水岭在于你把省下来的精力花在了哪。 很多人拿到强模型,图的是省事,恨不得三句话出个能跑的玩具就拿去发。 这哥们刚好反过来。 底层最吃工程量的硬活,他全甩给 AI 去啃。 为了还原手感,他先让 Opus 5.5 跑长程的物理仿真研究,观察真实战机的气动规律,再移植成浏览器能流畅跑的代码。 工程上直接开好几个 Agent 各自拉分支,联机排位、音效、抓 iPhone 闪退和加载卡顿同时跑,他自己在主干道上只干一件事:审代码合并。 那省下来的脑子,他拿去干嘛了? 全砸在了那些极其变态的体验细节上。 比如拿真实的座舱实拍录像去对齐 GoPro 鱼眼镜头的畸变,就为了让人一进去感觉像架了台真相机。 再比如专门去抠座舱盖的着色器,做出玻璃上的细微划痕迎着太阳反光、穿过云层时雨滴在玻璃上凝结滑落的效果。 说实话,看到这我真服了。 以前独立开发者一个人写完物理引擎和联机服务器,半条命都交代了,哪还有心思管玻璃上的划痕反不反光。 现在 AI 把 0 到 80 分的脏活累活包圆了,聪明人反而把 100% 的精力,全砸在了 90 到 99 分的审美和手感上。 这才是个人杠杆最狠的地方。 当然,这套打法能跑通有个前提:你自己得有兜底审分支的工程判断力,还得真见过好东西长什么样。如果脑子里连镜头畸变和光影质感的概念都没有,手底下挂十个 Opus 5.5,也只是在更高效地堆平庸代码。 刚去他网站(viperoneone.com)试飞了一把,看着阳光打在座舱玻璃划痕上的反光,我默默关掉了自己电脑里那个折腾了三天还在调登录框圆角的半成品项目。 reddit.com/r/ClaudeAI/commen…

ℏεsam avatar
⟳ Simon J.K. Pedersen retweeted
ℏεsam@Hesamation·Nov 3

BINANCE PUBLISHED MORE DETAILS ON HOW THE $6.34 BILLION HACK HAPPENED: >june: engineer A searches for a “popular ai coding tool” >downloads a malicious homebrew installer >binance security finds out >replaces the device >but engineers working on wallet infrastructure have been identified >september: engineer B gets contacted by a recruiter from a legitimate job platform >goes on a call with HR and later, CEO of a fake company >gets a coding assignment >downloads a malicious code repo on his WORK computer >malware runs, replaces a legit dev tool and goes undetected by security >october 20: hackers use engineer B’s computer and Binance’s internal ai tool to get access to engineer A and C’s computers >discover a token for wallet deployment on engineer A computer >october 29:log into engineer C computer >use engineer A’s stolen credentials >gain access to Binance’s wallet signing service so the hack wasn’t a spontaneous single security breach, but a sophisticated campaign over a few months, which also included social engineering. there’s particularly 3 main human mistakes throughout this process: 1. engineer A downloading a tampered installer from a search result rather than an official source 2. engineer B trusted a fake recruitment process 3. engineer B downloaded and ran an untrusted repo into his WORK laptop, which is honestly the most strange of them.

Binance @binance Nov 3 Article 29 October Binance Security Incident: Initial Findings and Response Update On 29 October 2026, Binance experienced a serious and highly sophisticated security incident. This should not have happened, and we apologise for the concern, disruption and uncertainty it caused our

CopyRebeldia avatar
⟳ Simon J.K. Pedersen retweeted
CopyRebeldia@CopyRebeldia·Oct 15

Este Fórmula 1 no es una grabación. Es un videojuego de 2014. Assetto Corsa con mods y DLSS 5, el renderizado neuronal de NVIDIA: la IA vuelve a dibujar cada fotograma en tiempo real. Hasta un diseñador profesional de videojuegos dudó si era real. Los estudios gastan cientos de millones persiguiendo el realismo. Aquí lo hicieron unos mods y una IA sobre un juego de hace 12 años. x.com/SimRacingEspor/status/…

Simon J.K. Pedersen avatar
Simon J.K. Pedersen@simped·Oct 15

Finally someone took the time to prove what I have tried to argue for a while. We need new computing concepts if we want to build AI that rivals the brain, the path we are on now is wrong.

demis hassabis mass-won the nobel prize in chemistry in 2024 for alphaFold. what nobody reported is what he said at a private dinner in london three weeks later. "we are mass-building the most inefficient thinking machines in the history of engineering. and we are calling it progress." a neuroscientist at UCL who was at that dinner spent the next 18 months turning that sentence into a 37-page quantitative proof. the paper was submitted to nature in august 2026. it was rejected. not because the science was wrong. because the implications were "too commercially sensitive to publish without further review." the paper leaked on a closed neuroscience slack on october 1st. Working Note: The Efficiency Gap Biological vs Silicon Intelligence - A Thermodynamic Analysis August 2026 here's what the paper proves - the human brain runs at 20 watts. the same power as a dim lightbulb. it processes approximately 10^16 operations per second. GPT-4 inference on an NVIDIA H100 cluster: 3,200,000 watts. it processes approximately 10^15 operations per second. the brain is 10x faster on 160,000x less energy. the efficiency ratio: brain = 5 × 10^14 ops/watt. H100 cluster = 3 × 10^8 ops/watt. the brain is 1,600,000x more energy-efficient than the best AI hardware on earth. now watch the historical curve [Fig. 1] - 2012 (AlexNet era): AI was 100,000,000x less efficient than the brain. 2017 (transformer era): 10,000,000x less efficient. 2023 (GPT-4 era): 1,600,000x less efficient. 2026 (current): 1,200,000x less efficient. at the current rate of improvement - 22% efficiency gain per year - AI hardware will match the brain's efficiency in 2089. sixty-three years from now. jensen huang's entire NVIDIA roadmap assumes exponential improvement. the paper proves the improvement is logarithmic. it's slowing down, not speeding up. the reason is thermodynamic. the landauer limit. rolf landauer proved in 1961 that erasing one bit of information requires a minimum energy of kT ln(2). at room temperature: 2.85 × 10^-21 joules. silicon transistors currently operate at approximately 1,000x above the landauer limit. every hardware generation closes that gap by roughly 15%. but the closer you get to the limit, the harder each improvement becomes. this is why moore's law died. the brain operates at 10x above the landauer limit. not 1,000x. ten. 3.8 billion years of evolution optimized the brain to within one order of magnitude of the physical limit of computation. NVIDIA has had 31 years and is still three orders of magnitude away. the paper asks one question in the conclusion: "what architectural principle allows biological neural networks to operate within 10x of the thermodynamic floor while artificial networks remain at 1,000x?" the answer: the brain doesn't separate memory and processing. every neuron is both storage and computation simultaneously. every synapse is both a wire and a processor. every computer built since von neumann's 1945 architecture separates memory from compute. data moves back and forth on a bus. the bus is the bottleneck. the bus is where the energy goes. NVIDIA's entire product line - every GPU, every tensor core, every NVLink - is an increasingly elaborate attempt to make the bus faster. not to eliminate it. faster. the brain has no bus. it never needed one. the paper's final line: "the companies that build the first commercially viable non-von-neumann inference chip will make the GPU obsolete within 5 years. the physics demands it." intel abandoned their neuromorphic chip program (loihi) in 2023. IBM shelved TrueNorth. the two companies closest to solving the problem quit. the $3.4 trillion NVIDIA valuation is a bet that the bus can be made fast enough forever. landauer proved it can't. 65 years ago. save this

0:28
Simon J.K. Pedersen avatar
Simon J.K. Pedersen@simped·Oct 5

Don't be like Boris ;)

Hi. I am a wheelchair seller. Buy my wheelchair. What's that?? You don't NEED a wheelchair?? No worries, this here is Boris. He will now break both your legs and for good measure, sever your lower spine so you become a paraplegic. See?! I was right! You need my wheelchair!

Simon J.K. Pedersen avatar
Simon J.K. Pedersen@simped·Oct 3

Wonder why you don't see any output gain from giving your developers Claude....

Claude Code just shipped mods... I made one that connects you to a multiplayer Doom server while Claude is busy. Every other player in the match is a person waiting for their Claude to finish.

0:25
BURKOV avatar
⟳ Simon J.K. Pedersen retweeted
BURKOV@burkov·Sep 17

IMO, Google doesn't try to compete with frontier LLMs because coding agents isn't what they want to do. It doesn't fit into their business model, and they don't feel they can disrupt the market with it the way they did with Search, Maps, Gmail, Android or even GCP. I think they see that a regular Google user doesn't notice any difference between the AI Overview quality (~GPT-3.5), the AI Mode quality (~GPT-4.0), and the frontier models. To be honest, I wouldn't see any difference between the models past o1 either if I didn't build with AI. So, Google is more interested in making GPT-4.0-level quality cheap than to beat it at any cost. Hence, the advancements in Flash and Flash Light and their Voice counterparts that would allow the users to talk to Google and that would be cheap enough to be paid for by the ads. Without a doubt, Gemini 3.7 Flash (as well as GPT-5.6 Luna) are the two best models for the money right now. I extensively tested all open-weight alternatives that are as expensive or cheaper, and none of them come close to these two models in terms of quality for money. I think it's the right strategy for Google. They don't lose casual users because those users don't see any reason to leave the Google ecosystem, Gemini is good enough as a voice interface to GCP and Gmail, and you can always easily distill your competitor's best model when that competitor reaches some sort of plateau, rather than try to beat all their intermediary versions for years for no practical reason.

Simon J.K. Pedersen avatar
Simon J.K. Pedersen@simped·Sep 18

Now we are talking. Make a general ai chip with this tech, and we can have AGI.

UCLA researchers built an image generation model that runs on “light” instead of GPUs. And it matches a 1.07 billion parameter diffusion model.. and generates each image in less than 1 nanosecond with basically zero computing power. It’s called “Optical Generative Model” Instead of pushing electrons through billions of transistors, it uses the physics of light to draw pictures. A digital encoder creates a noise pattern and imprints it onto a laser. The light then propagates physically through custom optical layers. The light itself does the calculating. There is no silicon bottleneck. It processes all-optically. And it uses almost zero compute power during the actual generation. It successfully created complex images, human faces, and Van Gogh-style art with the exact same quality as traditional AI. But the energy cost wasn't just reduced. It was practically eliminated. If this scales, the carbon footprint and computing limits of modern AI disappear overnight. We’ve spent years trying to optimize code to make models faster and cheaper. But the next massive leap in AI isn't going to be software.

Post attachment
Simon J.K. Pedersen avatar
Simon J.K. Pedersen@simped·Sep 16

Are you a meat proxy or slop grenade ;)

Lazy work used to mean too little output. Now, with AI, it often means too much and more work for everyone else. @tobi call it "slop grenades." A "Slop Grenade" is when you let AI produce the work and pass it on without adding any value (including checking it). Someone else has to wade through it, catch the mistakes, and clean up the mess. You save time and look productive but someone else pays for it.

1:16