https://xitter.cf/t.co/UjK3ZEETml

Joined January 2020
you know what else Jev helps classify? people who actually understand ML, and people who don't.
55
122
22
2,249
116,951
The "whiteboard defense:" I should be able to pull you aside at any moment and ask you to explain any customer-facing system you've shipped. You should be able to clearly explain how it works and defend the decisions you made. This is my benchmark for responsible AI usage. I don't expect line-level familiarity with the code. I don't care if you remember the exact function name or implementation detail. You may not even know it. I don't care. But if I ask "why did you do X instead of Y?", "what happens if this actor behaves maliciously?", "what data structure did you use here and why?", or "where does this fail?" you should be able to answer confidently. For PoCs, demos, experiments, whatever: I don't care. Generate 100% of it and understand none of it. Speed over quality every time in those specific scenarios. But if you're shipping customer-facing work, you can't be shipping things you don't understand at a high level.
234
1,018
196
9,497
488,180
injuly retweeted
DeepSource has been acquired by Harness. Sanket and I started DeepSource on one bet: every line of code shipped should meet a high bar for quality and security, without a human having to review it. Over the last eight years, we built DeepSource from the ground up: static analysis across 20+ languages, Autofix in 2020, before automated remediation was a category, code formatting, coverage, SCA and security, and eventually AI-powered code review and remediation. The idea never changed much. Find problems before code ships, and fix as many of them automatically as we can. With agents producing code faster than humans can review it, there is a lot more of that problem to solve now. I’m proud that everyone from open-source maintainers to enterprises relied on DeepSource to ship better code. More than anything, I’m proud of the team that built it. Harness is building the Autonomous SDLC Platform to catch risks in software before it ships, and our vision aligns perfectly with theirs. I'm excited that DeepSource is joining forces with them and continuing our work. deepsource.com/blog/deepsour…
21
6
3
92
4,354
Sachin Tendulkar has finally spoken on the student protests, and I had to read his note twice: once as a fan and once as an editor. The fan was disappointed, but the editor realised that this was a textbook specimen of the kind of writing that George Orwell warned about in his essay “Politics and the English Language”. The essay was written in 1946, but it might well have been written yesterday about this very statement. Orwell’s point was that when people want to hide something, they stop using plain words and reach instead for the passive voice, for fog, for phrases that sound like they mean something while meaning nothing at all. Consider the line, where students are disappointed that “their hard work hasn’t been rewarded”. Rewarded by whom, and denied by whom? There was a paper leak, which means somebody leaked it, some institution failed to prevent it, and an official presides over that institution. Yet none of them appears in the statement, and neither does the leak itself. A crime with identifiable custodians has been turned into something that simply happened to lakhs of children through nobody's fault. Then comes the list of those who must set things right: parents, teachers, friends, relatives, schools and administrators. If you read that list again slowly, you will notice it covers everyone in India except the people the students are actually protesting against, which is to say the NTA, the education ministry and the government. When everyone is responsible, nobody is, and that is not an accident of drafting. It was engineered to deflect blame. There is also a quiet sentence buried in the middle. “A society which prioritises outcomes over effort will seek shortcuts over meritocracy” is a line about those who take shortcuts, meaning the cheat. So a note on protests around paper leaks ends up gently lecturing the leaked-upon about honesty. His father’s lovely lesson—that failure is okay but cheating is not—is again aimed squarely at individuals, while the scandal and protests right now are about institutions that failed them. The rest of the statement is what Orwell called ready-made strips of words, glued together. Full of dreams and energy, fuel to our success, safeguard their aspirations. Nobody can disagree with any of it, which is precisely the point, for this is prose engineered so that no one in power will find a single sentence to object to. What an actual position would have looked like is a single sentence in the active voice: The government must find and punish those who leaked the paper, and fix the system that allowed it. He wrote 160+ words to avoid writing that one. Here’s Orwell’s full essay for those interested: bioinfo.uib.es/~joemiro/RecE…
128
1,000
66
3,600
282,873
Parents are asking their kids not to go for the protests in Delhi because they know who they voted for and what they are capable of.
5
150
1
1,419
12,553
injuly retweeted
please keep your water bottle caps tightened. BESCOM might cut the power to the entire city if the water spills.
2
27
831
Working in ML starts out as a math problem and very rapidly becomes a distributed systems problem
25
96
8
1,257
58,831
injuly retweeted
why is it so hard to make something customizable without ending up with settings like this
81
30
13
2,061
131,350
Tragically I am continuing to find that the most effective guardrail against slop is extremely talented engineers doing very thoughtful, human code review
87
204
46
3,349
165,857
injuly retweeted
69
2,508
56
23,338
1,204,035
It's crazy how exponentially complex a function needs to be to make gameplay actually fun vs simply functional. You can write a jump function with 2 lines of code, but it takes thousands more to create the 50+ micro-nuances of the jump to make it actually feel fun.
50
108
6
4,517
95,423
injuly retweeted
roon
@tszzlJan 22
there will be a cultural change at many software organizations soon where people declare bankruptcy on understanding the code they’re committing. sooner or later this will cause a systems failure that will be harder to debug than most, but will be resolved anyways
177
59
40
2,058
231,249
injuly retweeted
Maybe one day, topologists will have a 2nd joke
62
142
15
6,524
273,065
Thank you ⁦@AnirudhRowjee⁩ ⁦@amodm⁩ and all our friends @BengaluruSys⁩ ⁦for having us in Bengaluru! And ⁦@chaitybhandari⁩ for flying halfway across the world to join me on tour. An incredibly special day. ⁦1000x BLR!
3
15
4
99
13,094
"rewrite it in jax" is gonna be the next "rewrite it in rust"
14
5
4
300
21,345
Check out picoCAD by @johanpeitz, a tiny modeller for tiny models. Built on the PICO-8 platform, it focuses on the fun and essentials of lowpoly 3D modeling without the complexity of traditional tools. 💡 #pixelart #lowpoly #gamedev
28
620
27
6,401
209,875
I really sympathize with this one: openmymind.net/Im-Too-Dumb-F… I last hacked a Zig side project back in comfy 0.13 days, and now 0.16 confused me to the point I switched back to standard libraries I actually understand...
2
4
114
14,099
every debt is an asset if you never pay it back.
1
5
288
Man: why is the speed of light that? God: what Man: the speed of light in a vacuum is 299,792,458 meters per second… why? God: first of all, the speed of light is 1 dumbass
215
1,794
106
68,419
2,254,167
We're excited to open-source some of our work behind the scenes for the upcoming @autofixbot launch. The Narada model shows unprecedented accuracy for secret detection on public benchmarks. Exciting! xitter.cf/autofixbot/status/1977…
Meet Narada. A fine-tuned Llama3.2-3B-Instruct model that dramatically reduces false positives in secrets detection tools. The model achieves 97% precision with 96% recall on our evaluation set. It's fast enough for CI/CD, works with any regex-based tool, and is MIT-licensed.
2
5
569