Hasnain says:

So many really good quotes in here. Picking 2:

"The first step is to realize that the subway stops here. Up to this point in life, most of you have been rolling on train tracks. Elementary school, middle school, high school, college—it was always clear what the next stop was. In the process you've been trained to believe something that’s not true: that all of life is train tracks. And there are some jobs where you can make it stay like train tracks if you want, but really today is the last stop.

This fact is so terrifying that a lot of people try to remain in denial about it. (I certainly did.) But it's also exciting. You can go in any direction now."

"Now I have some good news: I'm almost done. I hate long speeches and I bet you do too. And frankly, if you can remember what I've told you so far, that will be enough. So let me remind you what I've told you: you've been able to go through life so far without steering much. If you want to, you can become more ambitious now, but to do that you have to start steering. You can't just drift. There’re a huge number of options, and you have to actively figure out which is the best for you. And the best way to do that is people. Find the interesting people. "

Hasnain says:

I wanted to quote the whole thing. Horrifying.

“It is too late for real justice in Gaza now. We can never bring back the dead children. We can’t erase what has happened. But it is not too late for accountability. The atrocities must be documented. The dead in Gaza must be properly counted so we know how many people have been murdered. The media must stop parroting the official death figure of more than 55,000 people being dead without putting this into context and noting that when you account for indirect deaths from starvation, disease, or cold, the real number of deaths is probably enormously higher.

If you have stayed quiet until now, telling yourself that all this is just far too complicated for you to speak up about, it is not too late to raise your voice. What is happening in Gaza is different from the horrors happening in Sudan and the Democratic Republic of Congo because, if you are in the west, it is happening in your name. It is happening with your tax money and with the help of your leaders. If you are in the US, your elected representatives have delivered a standing ovation for this genocide. We are all complicit. Although some of us are far more complicit than others.

So, again, think about what you want to say to future generations when they ask what you did at this very moment. Silence is not neutrality. And your silence will not be forgotten. As Martin Luther King Jr said: “In the end, we will remember not the words of our enemies, but the silence of our friends.””

Hasnain says:

I’ve always seen eg those articles on Buzzfeed and elsewhere that are basically of the form “someone posted X on the internet, and others responded Y and Z”

Did not realize there was a tech equivalent and that I’d get quoted for a random LinkedIn comment posted at midnight

… I do wish they had posted the (imo) more interesting parts of my comment though.

“Databricks software engineer Hasnain Lakhani wondered what he’d do when people simply lied about whether they’d used AI. (“Seems like an arms race,” he suggested, with projects needing tools to screen for AI.)”

Hasnain says:

Just came across this and bookmarking for future re reading, for.. reasons

“I think there is a tremendous opportunity for the open source community to really push the state of the art in coding evaluations. There's only so many benchmarks that I, personally, can create, but if everyone is making benchmarks I could eventually imagine a universe of benchmarks where you could curate the problems that are relevant to your work and quickly and cheaply judge models in this way: a Wikipedia of Coding Benchmarks.”

Hasnain says:

“But he didn’t give up. He teamed up with his longtime collaborator Tuomo Kuusi (opens a new tab), a mathematician at the University of Helsinki — “I’m almost married to him. I mean, how do you describe your best friend?” Armstrong said — along with Ahmed Bou-Rabee (opens a new tab), his postdoctoral researcher at Courant. The three mathematicians set out to fortify homogenization so that it would act like a rigorous version of the original renormalization argument.”

Hasnain says:

Unicode strikes again.

"When I flagged about this rather big omission to GitHub people, I got barely no responses at all and I get the feeling the impact of this flaw is not understood and acknowledged. Or perhaps they are all just too busy implementing the next AI feature we don’t want."

Hasnain says:

“Who has time to read all that? Who has the time for any of this? Technology is making our lives harder, not easier.

So I guess the next question is, "How do I fix this?" Like most people, I've been pulling back. Less time relying on algorithms to predict what I like and more time just making notes and lists in Obsidian. Any time I stumble across something that looks interesting or something I don't want to forget, I make a note of it so I can retrieve it later.”

Hasnain says:

“The GOOD version is good, because it avoids repeatedly re-evaluating condition, removes a branch from the hot loop, and potentially unlocks vectorization. This pattern works on a micro level and on a macro level — the good version is the architecture of TigerBeetle, where in the data plane we operate on batches of objects at the same time, to amortize the cost of decision making in the control plane.”

Hasnain says:

“It’s true that a lot of open source projects really hate AI code. There’s several objections, but the biggest one is that users who don’t understand their own lack of competence spam the projects with time-wasting AI garbage. The Curl project banned AI-generated security reports because they were getting flooded with automated AI-generated “bug bounty” requests. [LinkedIn]

More broadly, the very hardest problem in open source is not code, it’s people — how to work with others. Some AI users just don’t understand the level they simply aren’t working at.

One user of the LLVM compiler complained that his AI-generated pull requests were not being taken seriously — by a compiler project, where correct computer science and knowing precisely what the heck you’re doing is quite important.

The user considered it was the unpaid volunteer coders’ “job” to take his AI submissions seriously. He even filed a code of conduct complaint with the project against the developers. This was not upheld. So he proclaimed the project corrupt. [GitHub; Seylaw, archive]

“

Hasnain says:

This was a pretty motivating read! I’ve been doing this type of work recently for learning purposes and having an MCP for eg pwndbg would make this so much easier for a noob like me

(Now I must ensure I don’t get distracted by yet another rabbit hole on side projects…)

“What Does This Mean In Practice?

Let me walk you through what this enables:

Natural language crash analysis: "Why is this application crashing with an access violation at this address?" (Instead of: "What the $%#@ is this heap corruption!?")

Contextual debugging: "Show me the stack trace for thread 5 and explain what each function is doing based on the symbols." (Instead of staring at call stacks like they're ancient hieroglyphics)

Root cause identification: "What's causing this null pointer dereference and where should I look in the code to fix it?" (Instead of playing detective with memory addresses)

Instead of typing obscure commands like !analyze -v followed by a series of manual investigations, you simply ask questions in plain language, and the AI interprets the crash data for you. It's like having a WinDBG expert whispering in your ear, except it doesn't get annoyed when you ask the same question five times.

“