Hasnain says:

“If you can justify an AI coding subscription, you can justify buying the best tool for the job.”

Hasnain says:

“The net result of leaning extremely heavily into type checkers and unit tests, is that the LLM can quickly loop on itself to fix these problem, resulting in eliminating a whole slew of issues with vibe coded output. “

Hasnain says:

The only thing that surprises me about this quote is that first word

"Surprisingly, I think that the fact that I need to plan my features carefully before rushing into implementation is making me a better developer overall. This happens simply because it forces me to document the implementation and think it through before jumping into code. I also find myself explaining my reasoning more clearly because I have to write it down for the AI, whereas with colleagues I would typically discuss things in person or via video call."

Hasnain says:

“Looking further ahead, our long-term solution involves building a new, enhanced traffic management system. This system will allot network resources on a per-customer basis, creating a budget that, once exceeded, will prevent a customer's traffic from degrading the service for anyone else on the platform. This system will also allow us to automate many of the manual actions that were taken to attempt to remediate the congestion seen during this incident.”

Hasnain says:

wut

“At the meeting, some people had reasonable explanations for not getting their AI assistant accounts set up during the week, like being on vacation, Armstrong said.
“I jumped on this call on Saturday and there were a couple people that had not done it. Some of them had a good reason, because they were just getting back from some trip or something, and some of them didn’t [have a good reason]. And they got fired.””

Hasnain says:

“In recap. What you just built was a coding agent.

Perhaps you don't want to create a coding agent. What if you're in the data engineering profession? What would that look like? Think about all of the activities that you do day-to-day, where having the capability to automate using these primitives could be handy or valuable to your employer.

Your current workers are going to take your job, not AI.

If you're concerned about AI, the answer is straightforward: just invest in yourself. It really is that simple. This year is a particularly challenging time to be asleep at the wheel when it comes to personal development.”

Hasnain says:

“The beautiful thing about getting comfortable with repair is that it actually makes you better as a manager. When you know you can fix things when they go wrong, you’re more willing to make decisions, have difficult conversations, and take reasonable risks. You stop being paralyzed by perfectionism because you know that most mistakes, while serious, create opportunities for growth and stronger relationships when handled well.”

Hasnain says:

“I don’t know yet if the implementations of this yet are good enough to use at scale. Maybe they’re slow or maybe the bugs aren’t ironed out yet. But obviously it’s possible to build a system that takes an arbitrary declarative, stateless query and does this sort of static analysis and incremental computation behind the scenes, and it should be possible to make it fast and reliable. And if you can make it good, it’s obviously extremely useful. You get to cut out all of that application code that’s dealing with keeping things in sync and make dealing with the stateful data updates someone else’s job. And ideally you can change the performance characteristics (eg: how much to store in memory, whether the result needs to be updated immediately or can be deferred, etc) without rewriting all your code that actually computes the answer.

It’s too good of an idea for it to not succeed. Certainly if I was in charge of databases at AWS, this would be a major tentpole for my roadmap! I figure that a decade from now, most database systems will have a version of this built in.

I can’t wait until they do.”

Hasnain says:

“This new result is still just a lower limit on BB(6) — the true value could be even higher. Busy beaver hunters don’t expect to have a definitive answer anytime soon. The first sign of trouble was a monstrous six-rule Turing machine that the team has named Antihydra (opens a new tab), discovered by mxdys last year.

Antihydra almost certainly never halts. But researchers haven’t been able to prove it. And there’s a good reason for that: A busy beaver hunter who goes by Racheline has shown that the question of whether Antihydra halts is closely related to a famous unsolved problem in mathematics called the Collatz conjecture. Since then, the team has discovered many other six-rule machines with similar characteristics. Slaying the Antihydra and its brethren will require conceptual breakthroughs in pure mathematics.”

Hasnain says:

Lots of fun internal insights in this one.

“The main takeaway, again, is to keep things simple. Extreme scaffolding frameworks will hurt more than help you. Claude Code really made me believe that an "agent" can be simple and yet extremely powerful. We've incorporated a bunch of these lessons into MinusX, and are continuing to incorporate more.”