Aug 16, 2026

Focus and Speed

This is your #1 most important challenge. If you master focus, you will be in control of your world. If you don’t, it will control you. And most importantly, be one of the few that stays in the shed to practice, while everyone else is surfing the net, flirting, and watching TV.

Berklee classes set a pace the average student can keep. If you want to be above average, you must push yourself to do more than required.

July 31, 2026

Should you use AI for a task?

At work, if your job is to move a bunch of heavy things from one side of the room to another, you should use whatever assistive tech you have on hand: a wagon, a forklift… even an AI-powered robot. But at the gym, it makes no sense for that robot to lift weights for you. The point of weightlifting isn’t to move heavy things across the room; it’s to actually lift those heavy things.

The writing assignments I give my students are gym tasks, not work tasks. I ask them to write policy memos not because the world needs more policy memos. I assign them because the very act of writing, which includes thinking and outlining and drafting and editing, making and criticizing and revising arguments, will help develop the critical thinking skills they will need in their future careers. And without this constant mental exercise, those skills will atrophy. Employers are already noticing.

What the students miss is that their initial discomfort is a normal and healthy stage of writing, and not something to quickly get beyond. The very act of struggling with how to express what they think is an important part of the process.

July 27, 2026

Design is all about tradeoff

Compromise is neither good nor bad, it’s something we do every day. It’s decision making. Prioritizing. Deciding that one thing is more important than another. It’s finding the right balance between two competing desires.

Having an opinionated set of tradeoffs exposes your approach to a set of weaknesses. The more you tip the scale on one side, the weaker something else will be. That’s okay! Making those difficult choices is what people pay you for. You should be proud of your compromises.


July 23, 2026

Job interview could be an attack

Hãy cần cẩn thận khi tham gia phỏng vấn cho một công ty ngẫu nhiên nào đó mà yêu cầu làm “take-home assignment”. Trong quá trình phỏng vấn, công ty sẽ yêu cầu bạn làm 1 bài tập nhỏ và gửi lại kết quả, thường sẽ là debug/implement một feature/code nào đó. Trong trường hợp của blog này, tác giả được yêu cầu clone 1 github repo về. Nhìn sơ qua thì repo này khá giống một repo backend thông thường. Tuy nhiên điều đáng nghi là repo có chứa nhiều git hooks, và một trong các đoạn pre-commit hooks sẽ chạy đoạn mã download mã độc online rồi chạy trên máy của mình.

Sau khi đào sâu hơn, tác giả thấy rằng có nhiều cách khách nữa kể kích hoạt mã độc, ví dụ như mở repo bằng Vscode.

After realizing this was a widespread campaign, I did a bit more digging and found that people are getting different variations of this attack. Some folks received a zip file containing a .vscode folder. Inside, the attackers hid commands configured to run as soon as the directory is opened in VSCode (launch commands).


July 20, 2026

The impact of AI on decision-making

Several of my peers now “AI-wash” their work, meaning that even when they can perfectly competently execute on their jobs to the satisfaction of their management teams, said managers are unhappy if the engineers haven’t used AI in the work… so now they’re lying about using LLMs even in contexts where their professional judgement is that they aren’t the appropriate tool.


Theo Artificial Analysis, để tối ưu chi phí cho coding tasks, thì nên sử dụng GPT5.6 Luna xhigh/max rẻ hơn nhiều so với GPT5.6 Sol medium mà vẫn cho kết quả tương đương.
Thông minh tăng dần: Opus4.8 max ~~ GPT5.6 luna xhigh -> luna max -> tera max -> sol high


July 13, 2026

Reverse information paradox

You essentially pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful. The better you want the model to perform, the more of that knowledge you have to feed it!

Over time, the information asymmetry becomes increasingly skewed. The seller learns more and more about you as you use what you purchased, while you learn very little about what the seller is learning in return.


July 6, 2026

Dan Luu on AI-coding

A lot of getting value out of agents seems to be having some kind of understanding of their failure modes and then working around them. Earlier in the post, we noted that people aren’t really going to be left behind if they’re not using coding agents now and get started in M months because they can be at most M months behind, but will likely be much less behind than that because of how quickly things are changing. A major reason for this is, AFAICT, a lot of the skill involved in using agents is working around their failures. Of course AI labs want to fix these failures, which means new releases are designed to obsolete these skills as much as possible. This is possible to observe directly, in some failure modes that were quite common a year ago are now much rarer.