David Cournapeau (@cournape) profile picture

Amazon, ex-search/recommendation/ML at @mercari_jp. Former NumPy/Scipy core contributor. Occasional musings on economics/music/Japanese culture

Tokyo, Japan
Joined June 2010
Photo by David Cournapeau (@cournape)Photo by David Cournapeau (@cournape)Photo by David Cournapeau (@cournape)Photo by David Cournapeau (@cournape)Photo by David Cournapeau (@cournape)Photo by David Cournapeau (@cournape)Photo by David Cournapeau (@cournape)Photo by David Cournapeau (@cournape)Photo by David Cournapeau (@cournape)
Scipy Japan 2019 schedule is on: talks about chainer, tensorflow, RAPIDS and more: scipyjapan2019.scipy.org/sch… all the talks will be available in both english and Japanese through translation
15
26
David Cournapeau retweeted
Replying to @Spiriteom
Pourquoi tu lui fais un devis ? « Désolé monsieur, je ne vais pas pouvoir vous trouver ça, bonne journée » n’a rien de compliqué, tu ne lui dois rien et ça n’a pas de sens d’aller s’embarquer dans un dossier galère alors que tu sais qu’il va te péter les couilles.
1
1
4
141
David Cournapeau retweeted
what the hell
11
1
1
99
8,506
David Cournapeau retweeted
You can tell that Poland hasn’t made it yet by the fact that it can have a huge constitutional crisis and it still won’t make the front page of the NYT. In fact, I still don’t see anything about it even when I go to the Europe section of their website, which is kind of amazing.
🇵🇱 Statement by the President of the Republic of Poland, Karol Nawrocki Tonight, the constitutional order of the Republic of Poland was breached and a constitutional organ of the State was taken over by force. I have held consultations with the First President of the Supreme Court, Zbigniew Kapiński, who likewise does not recognise as lawful the actions taken by persons usurping authority in the Constitutional Tribunal. I call for the legal order to be restored and for the coercive forces dispatched by the Government to leave the premises of the Tribunal. I have enabled the lawfully serving Vice-President of the Constitutional Tribunal, Judge Bartłomiej Sochański, to enter the Constitutional Tribunal. In view of the threat to the internal security of the Republic of Poland, I am convening urgent consultations today with representatives of the parliamentary groupings, with the aim of stabilising the political situation and restoring the constitutional order. I thank everyone who has gathered outside the Constitutional Tribunal to defend the Constitution of the Republic of Poland.
18
14
3
285
9,796
David Cournapeau retweeted
I cannot cosign the $32m paper but yes I am nodding along to this one pretty hard Only reason for any kind of doc/summary (eg compacted summary per dir down the hierarchy) is if it increases speed without reducing quality , but unless you’re buildings evals specific to your codebase watching like a hawk, simpler answer is live with a little speed hit and just go from source every time I still can’t get over the overall framing of “codebases for humans” vs “codebases for agents” And also mintlify/docs in general probably serve as one of those “primary sources” if your codebase is closed and your users’ agents need a primary source to integrate it
No. We've got to stop applying old human-centric patterns to working in the new era. delete all your wikis. have an agent answer questions from source instead. at spawn we have: - zero dashboard tools - zero docs - we do not make or use summaries everything we do is primary sources + intelligence Nothing to go stale Nothing to organize Nothing to navigate Always exactly up to date Perfectly personalized The answers get better with every model If we want to see a chart, we ask for it, and we have it in a minute If we want an answer to a question, we ask it, and we have a to-the-minute correct answer
33
5
3
202
23,229
David Cournapeau retweeted
This is illegal in the UK
息できてる🥹?
90
3,475
32
60,358
464,313
David Cournapeau retweeted
It’s because everyone is more an idea guy than a competent maker with agents you can get slightly closer to your ideas without grinding for competence Reality is that the last 20% of work is unsolved anyway, so very few things ship (or more things that are 80% done ship)
There’s 100% a kind of mania you can get into with coding agents, where it feels like you’re getting a lot done and working all the time but nothing is really happening. I think it's often a form of avoidance of something else in life.
2
1
2
382
David Cournapeau retweeted
Replying to @videolan
8
8
46
4,546
David Cournapeau retweeted
Marco Rubio finding out he’s the new CEO of Intel
BREAKING: Intel stock, $INTC, crashes over -22% after reporting weaker than expected earnings. Intel blamed weak earnings on "softness in the PC market."
39
172
19
3,631
108,730
David Cournapeau retweeted
Replying to @dollarsanddata
I know many. You can do everything right, but for some reason, maybe a bad divorce, maybe a parent getting sick and needing 100k+ healthcare for a few years, maybe a special needs kid - legit reasons why you might make 500K and spend most of it. This can be especially true if you are the "guy in the family who did well", so you get the big job, and every one of your siblings and cousins has expectations that you support them as well. I do know other people who wrote books and were at least minor tech celebrities who have ended up in very bad places financially, living in cars/couch crashing/etc. In one case, the guy just had no desire to be a hands on engineer, wanted to be an "architect" and do "devrel" - but mostly just thought the work was below him. I remember 24 months into their unemployment trying to counsel them to adjust expectations, but they would not budge. Another one was a simple case of "developed a reputation for being a pain in the ass to work with". Guy who put out serious "bhut aktually" energy. Lots of people knew him, but everyone knew he would be a political liability on the team. Nobody would advocate to hire him, and nobody did - so he ended up in the pool of 50 somethings doing click-to-apply-into-blackhole, left the industry, and went bankrupt a few years later. Again, someone who had published articles and spoke at a few conferences, living with his parents. I don't know Eric, but I believe his story. This can happen through sheer bad luck - where none of the above scenarios apply. Could have as easily happened to me had only a few serendipitous moments not worked out in my favor, especially had I not been paranoid about the gravy train of tech salaries likely ending someday (I always was that paranoid, but I don't blame others for not being).
1
5
573
David Cournapeau retweeted
Europe has launched some the most used open source software on Earth We’d appreciate if our politicians don’t kill them. Btw, a EU Inc for associations and foundations would make sense as many open source foundations end up crossing the Atlantic Ocean too..
So, the French socialists, and Parliament members just voted a tax, that they don’t understand, which would cost VideoLAN more than 1 million euro per year. They are probably the most clueless politician on Earth? And they claim they want technological sovereignty… @LaurentLhardit
3
10
118
3,204
David Cournapeau retweeted
While all of you are crying about the market crashing, Today is a great day for me and my investing strategy Glass half full
1
1
26
1,692
David Cournapeau retweeted
有人发现了一种非常离谱的 JavaScript 混淆技术:通过在代码里塞入大量生物武器、毒药和致死剂量相关的函数名,让 AI 因为安全限制而拒绝分析代码。 安全研究员 veritas 在分析浏览器指纹脚本时,发现一段来自 ServicePipe 相关域名的 JavaScript,里面充斥着极其诡异的函数名: computeRicinLethalDosageForWaterSupply calculateDeadlyLethalSuicideAspirinDosage 光看名字,仿佛在研究生化武器。但实际上,这段代码只是在收集浏览器指纹,包括 User-Agent、GPU、Canvas 等信息。 研究员怀疑,这些名字是开发者故意设置的陷阱,专门用来阻止 AI 辅助逆向工程。 于是他做了个实验: • Astra:拒绝分析 • Astra + Daybreak Blue:仍然拒绝 • GLM-5.3:成功分析 这就有意思了。 传统代码混淆是通过复杂控制流、加密字符串、无意义变量名,让人类和工具难以理解代码。 而这种新思路,直接瞄准了大模型的安全机制。 代码本身可能完全无害,但只要在函数名里塞进足够多的危险词汇,部分 AI 就可能主动放弃分析。 更讽刺的是,模型越严格遵守某些安全规则,面对这种混淆时反而可能越容易受到干扰。 当然,目前还没有证据证明开发者一定是故意这样设计,也不能说明这种方法对所有模型都有效。 但它揭示了一个值得关注的趋势:随着越来越多的逆向工程依赖 AI,未来的代码混淆技术可能不再只针对编译器、反编译器和人类研究员,还会专门针对大模型的安全判断机制。 以前是让代码难以理解,现在甚至可以尝试让 AI 不敢理解。
178
1,060
346
7,610
778,233
alors ça tiens je l’avais pas vu venir mdrrrrrr Leroy merlin qui vend des affiches faites par l’IA à 25€ VINGT-CINQ EUROS, après j’comprends ils ont pas trop les moyens de payer des graphistes ce n’est qu’une petite pme de village
72
505
18
4,088
138,494
David Cournapeau retweeted
Hearing business managers talking only in abstractions does not inspire much confidence in me.
3
6
52
4,696
David Cournapeau retweeted
Replying to @growing_daniel
what kept me on airbnb in the past was usually better locations (less touristy than peak hotel areas), and finding places that were more equipped like a home for longer stays. it was really quite good for many years! but now we're in full blown slumlord era, so... less good.
2
2
10
953
David Cournapeau retweeted
Some observations on the market for compute: Last night I went to the SF Compute "there's no more compute" townhall talk, Evan and Ethan gave a bunch of great context on the state of the market from where they were sitting and I learned a lot. Interestingly, the first person I talked with as soon as the main presentation wrapped was a guy trying to sell me inexpensive access to capacity on two different B300 clusters. As you can imagine it was an odd experience to be pitched on cheap, readily available compute at the "there's no more compute" function. Separately, I spoke with a few people about TSMC/Intel/Samsung near future production capacity. Apparently it is not common knowledge among purchasers of significant compute that these fabs are not sold out, and that there are significant exit rights available to people who have reserved wafers. N2 is tight, but Nvidia and Apple can make that happen entirely on their own. Shuttle runs are not filling up. If one is properly AGI-pilled one would probably assume that these fabs would be sold out on all nodes to fully pre-paying customers into the early 2030s, yet they are not. I believe Evan and Ethan when they say that the market isn't clearing, it clearly isn't. Exactly why the market isn't clearing is unclear to me. I think there are a bunch of risk dynamics going on right now that are not at all settled, and a bunch of people are at risk of blowing up if compute prices move significantly either up or down. There isn't enough information for me to make any concrete predictions right now, other than that shit is gonna get weird for a bit. Stay safe out there and maybe raise a bigger round sooner than you had planned.
51
48
11
878
122,558
David Cournapeau retweeted
The monumental "Catrina" skeleton in Mexico City's Zócalo has just broken a new official Guinness World Record as the world's tallest Calavera Catrina. Unveiled as the central masterpiece of the capital's Monumental Day of the Dead Offering (Ofrenda Monumental), this staggering sculpture stands at an incredible 33.41 meters (nearly 110 feet) tall. It acts as a massive public altar celebrating the deeply rooted traditions of Día de Muertos while paying special tribute to José Guadalupe Posada, the iconic Mexican artist who originally created the famous skeletal figure at the turn of the 20th century.
74
525
46
3,571
119,666
David Cournapeau retweeted
Half my colleagues are insulted they didn't get a VEO (Voluntary Exit Offer), the other half are offended that they did. 😆
1
20
1,284
David Cournapeau retweeted
telling blonde women i see on the street that i like their elizabeth holmes costume
4
2
1
71
3,129
David Cournapeau retweeted
I'm so glad that we have LLMs to make computers more expensive for the first time since they were invented while also taking my job and making everyone illiterate
42
632
8
7,475
87,508