Pandu Pradhana (@phynx17) header
Pandu Pradhana (@phynx17) profile picture

A programmer for a living, a husband to an Angel, likes to play with lots of data.

Joined June 2009
Photo by Pandu Pradhana (@phynx17)Photo by Pandu Pradhana (@phynx17)Photo by Pandu Pradhana (@phynx17)Photo by Pandu Pradhana (@phynx17)Photo by Pandu Pradhana (@phynx17)Photo by Pandu Pradhana (@phynx17)Photo by Pandu Pradhana (@phynx17)Photo by Pandu Pradhana (@phynx17)Photo by Pandu Pradhana (@phynx17)Photo by Pandu Pradhana (@phynx17)
🛡️ CVE pages are live in Zero Wire on Cloudflare Radar! Every published CVE now has its own page, with what you need to decide how urgently to act. This includes: - Exploitation signals: CISA KEV, known ransomware use, SSVC and EPSS. - Affected products and recommended actions. - Every CVSS score, with each metric explained. - Follow new CVEs via RSS, Atom or JSON, filtered by severity 🔎 Explore CVEs: radar.cloudflare.com/zero-wi… 👀 Read the changelog: developers.cloudflare.com/ch…
36
246
28
1,431
182,190
Pandu Pradhana retweeted
Replying to @paulg
@paulg blocked me years ago, so he did hear of me at some point. And it's true I don't much matter, none of us really do. I think my track record is reasonable. For instance, I helped get a multi-billion Federal Reserve subsidy for large banks cut and pass a law auditing emergency lending facilities at the Fed. I warned of Covid and supply chain shocks before anyone knew what was going on in Feb 2020, I published key details about a fire truck oligopoly that has led to dozens of antitrust suits, my book helped lead to the revival of antitrust law and big tech investigations. On the negative side, I supported the war in Iraq, I failed in my advocacy to end the foreclosure crisis, and most of the antitrust cases against big tech have gotten hung up in the courts. And I'm kind of a jerk on X. The response is likely to be 'feh advocacy what a LOSER.' Fine. You don't have to respect politics, advocacy, or law. But I don't have to respect Elon Musk or the overstated importance of financiers who think they are builders. Musk is a PR impresario, when he actually had to do something difficult in a context he didn't control - DOGE - he failed completely. He's good at some things, not others. But he, like most of the Silicon Valley elite set, is just a normal person. You guys elevating him into a religious icon makes your whole builder schtick a totally laughable garbage ideology. I'm sorry Mommy Schumpeter lied to you and you surrounded yourself with yes men. So I'll let you in on a little secret. You guys aren't special, you aren't especially hard workers, you aren't going to live forever, and if you didn't exist our society would be just fine.
220
37
14
347
68,662
Pandu Pradhana retweeted
The $NVDA news just keeps getting worse. - Entire administration is turning on him. Rubio likely leaked the initial reports and Bessent went on Fox yesterday and demanded a new CEO and answers from the Nvidia board. - New investigation today found empty server racks in Indonesia from Megaspeed. An affair could be forgiven, selling to China is going to be harder for the market to forgive if it was intentional. - Semianalysis just cut their 2028 capex forecast by 28%, indicating that capex growth has peaked. This really hurts NVDA. - Jensen now has to deal with multiple investigations and the company, going into earnings, might say anything that gets interpreted as bearish when it comes to guidance/growth/margins even if it is slightly off from expectations. - If the new CEO is going to cut most of the circular financing that Jensen did, may not be as easy for them to hit numbers when some of these smaller AI companies don’t have the funding to buy their chips. I know people are saying that the valuation has discounted all of this from $240, sure, but if this stock isn’t trading on valuation and more on sentiment, if the sentiment becomes that it’s just too hard to own…thats where you get the ugliness continuing. Just becoming harder to measure the risks that are piling up.
540
411
65
6,154
1,278,219
‼️ UPDATE: Microsoft says last week's global Microsoft 365 outage came from a fix it rolled out in July, which hid a bug for over three months. The bug froze the gateway servers in front of Exchange Online email and other Microsoft 365 services under heavy load. Lab tests on weaker machines missed it. Hours earlier, Teams had moved its traffic onto those servers, while nearly 14% were already offline. Microsoft's automatic failover then spread it. In under two hours, more than 80% of the servers stopped responding. Servers that came back were flooded by waiting traffic and froze again.
31
162
18
1,148
101,290
Pandu Pradhana retweeted
I blocked Artex without lifting a finger, it took 3 hours to get da where one of my shitty batch files would have worked in less than 60 seconds. This blog should be a good read: (But not for me tonight I need to sleep) 🫡✌️🤙
7
26
1
327
42,033
Pandu Pradhana retweeted
“It takes 20 years to build a reputation and five minutes to ruin it.” — Warren Buffett
163
1,337
123
12,744
235,097
‼️Update: Microsoft confirmed that its recent fixes haven't worked as expected. Engineers are exploring other ways to increase capacity and address ongoing traffic imbalances. Administrators can track updates in the Microsoft 365 admin center under EX1550543 and MO1550821. Details: cybersecuritynews.com/massiv…
🔻 Massive Microsoft 365 Outage Impacts Outlook, Teams, SharePoint, Copilot, & Defender XDR users Details: cybersecuritynews.com/massiv… A major Microsoft 365 outage disrupted cloud service access on November 12 and continued into November 13, 2026, as Microsoft worked to restore failing infrastructure. Microsoft confirmed that connectivity problems could affect users in any region, expanding the scope beyond its initial investigation into Exchange Online mailbox access in the Americas. The reported impact spans Outlook, Teams, SharePoint, Copilot, and Microsoft Defender XDR, although the exact service and feature restrictions remain subject to Microsoft’s updated advisory. Administrators should check incident MO1550821 for that breakdown. #Microsoftoutage #Teams #Exchange #cybersecurityNews
83
604
68
1,666
248,629
Pandu Pradhana retweeted
I have a major result to share. I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China. This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model. Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves. I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems. I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used. It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20. These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs. I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo…. There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.
1,822
6,424
2,210
41,254
6,786,288
With the recent wave of vulnerability disclosures, it’s a good time to take a closer look at your own code. Turn your coding agent into a team of security auditors with the open-source skill that seeded Cloudflare’s vulnerability discovery harness. The 6 phase approach: 1. Map the codebase 2. Hunt for vulnerabilities 3. Challenge each candidate 4. Record the findings 5. Verify the findings 6. Report what’s confirmed, rejected, or still needs validation Start your hunt here: github.com/cloudflare/securi…
236
1,581
294
18,150
2,459,013
Pandu Pradhana retweeted
Earlier this week Tristan published his top YouTube channels: thume.ca/2026/11/01/youtube-… Which looked excellent so clearly I had to turn it into TristanTube: karpathy.ai/tristan-tube/ 47K super nerdy videos +TV mode for laid back watching. (One-shotted with Claude Code in 3 minutes)
Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5: hackertube.org/ - Excellent recs without any logins (state is kept in local browser session) - Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads - Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version - HackerTube Longs ✊ A bit more on the process: I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing. Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.
739
794
243
9,439
1,086,018
🚨 速報:15カ国で同時に、銀行へのハッキングと口座からの窃盗が報告されている。 新型コロナも1週目はこんな様子だった。散発的な見出し。小さな数字。誰も危機とは呼ばなかった。 1/ 日本:銀行を装った偽の電話が、山形の企業から9億1200万円(約610万ドル)超をだまし取った。10月にはベトナムとの合同事件になった。現地で7人逮捕。 2/ 台湾:1日で永豊銀行の顧客418人から4176万台湾ドル(約130万ドル)が引き出された。ログインしていなかった顧客まで。返金済み。 3/ ベトナム:ハッカーは何カ月もの間、銀行のシステム内に潜んでいた。そして一夜にして銀行自身の資金を奪った。 4/ 韓国:約1カ月で金融会社37社が侵害された。約47万3000件の記録が流出。約580社が調査された。 5/ ネパール:中央銀行が各銀行にウェブサイトをオフラインにするよう指示した。顧客22人が偽のKYC電話で498万7000ネパール・ルピーを失った。 6/ インドネシア:ジャンビ銀行の6,609口座が被害に。1448億2000万ルピア(約880万ドル)が流出した。 7/ シンガポール:偽の署名入り書簡によって1280万ドルがカンボジアへ送金された。全額回収。 8/ マレーシア:2つの銀行にある王族の口座から198万5000リンギットが抜き取られた(民事訴訟)。 9/ スリランカ:財務省がメール詐欺で250万ドルを失った。いまだ回収されていない。 10/ ニュージーランド:偽の警官が22人の銀行口座を乗っ取った。19万8000NZドルが消えた。 11/ カナダ:カナダロイヤル銀行(RBC)の元従業員が、組織犯罪のために偽口座43件を開設したとして有罪に。 12/ メキシコ:ある犯罪集団がSPEI決済システムに対してAIツールを使った。 13/ ブラジル:中央銀行で9カ月間に過去最多の99件のサイバーインシデント、うち68件が詐欺。 14/ 米国:ダラスで14人が起訴:SIMスワップと買収された内部関係者が、3億7300万ドル超を保有する口座を標的にした。フロリダ州の企業は3日間に15回の電信送金で650万ドルを失い、400万ドル超がいまだ行方不明。 15/ フランス:クレディ・ミュチュエル・アルケアが取引先を通じて約1億3300万ドルを失った。15件の中で最大の損失。 では、視野を広げてみよう。 窃盗犯はアプリ、取引先、偽の電話、買収された職員を通じて侵入する。ブラジルは9カ月で2025年の年間件数を上回った。これはまだ始まりにすぎない。 新型コロナは街を封鎖した。銀行の封鎖を命じた政府はない。窃盗犯にはそもそも必要なかった。まだ誰もこれを危機とは呼んでいない。だが日本の警察はすでに、被害者が詐欺師に2度支払う日に備えている。 来週から、日本の警察庁は詐欺口座のデータを807の金融機関と共有する。日本のネットと電話による詐欺の被害額は1月から9月で2610億円(約17億ドル)に達し、2025年の過去最多を上回るペースだ。 ニュースサイトはこうした窃盗を1カ国ずつしか報じず、15カ国すべてをまとめて報じることは決してない。銀行に預金があるすべての人に送ってほしい。
117
2,249
239
6,165
1,034,742
Pandu Pradhana retweeted
Hey everyone, we have heard your concerns about Proton’s AI strategy and culture and after internal discussions, we want to share a few things with you. First, Proton does not install tracking software on employees' laptops to measure how much AI-generated code they produce. This rumor stemmed from a misunderstanding about how MDM software works, and it's not true. Second, our use of AI will never be at the expense of security. In the past year, contrary to what has been alleged, additional layers of security review have been added to meet the threat posed by hostile AIs scanning Proton's open-source code for vulnerabilities. Our code is open source, and its security remains open for all to verify. We want to prove this to you, so in the interest of transparency, we will lift the hood on how we use AI internally at Proton. This includes publicly releasing and open sourcing several of the internal tools that we use, including more advanced versions of Lumo with greater capabilities, which have never been made available outside of the company. We will also publish how code reviews and security reviews at Proton are done, and detail our AI policies, including where AI is permitted, and where it is not, and our rationale. Instead of just being able to see the final open-source code that we produce to verify security, you will also be able to assess how that code was produced. Lastly, we are not re-interviewing our staff about AI usage, and there are no job losses due to AI re-interviews. You can hold us to these commitments. We're aware that we haven't lived up to the high standard you ought to hold us to, and for that we can only apologize, we are sorry.
688
380
153
5,957
614,280
Pandu Pradhana retweeted
OpenAI 把整套 Codex Harness 开源了! 除了 CLI,还有 Skill 和运行环境,想学习 Agent 构建的朋友们有福啦! github.com/openai/codex
71
392
27
3,728
302,160
Pandu Pradhana retweeted
Bullish on Andy Yen and Proton. He gets it. The culture war, or rather the loss of identity some engineers feel as AI redefines our jobs is difficult; I sympathize. Also understand that he runs a business and might has limited patience when we now had two years to process, learn, change and grow.
Replying to @josephfcox
Here is Proton CTO's AI manifesto, that I obtained multiple copies of 404media.co/manifestos-fear-…
118
98
27
1,701
225,470
Pandu Pradhana retweeted
I am shocked, absolutely shocked, that Anthropic ARR has likely gone vertical in October after they cut token prices. Except the exact same thing happened for the entire AI market, which is why GPU rental prices have surged while token prices declined. $100b plus in year-end ARR looks possible again.
276
511
127
5,426
814,255