weekend ai reads for 2025-06-06

The idea is to identify opportunities to buy mature, people-intensive outfits like law firms and other professional services firms, help them scale through AI, then use the improved margins to acquire other such enterprises and repeat the process.

But the fundamental problem is that LLMs don’t get better over time the way a human would. The lack of continual learning is a huge huge problem. The LLM baseline at many tasks might be higher than an average human’s. But there’s no way to give a model high level feedback. You’re stuck with the abilities you get out of the box.

I’ve seen this firsthand: Some of the most brilliant founders, executives, and creative professionals I’ve worked with weren’t stuck because they lacked intelligence or strategy. They were stuck because they didn’t know how to deal with their emotions.

First, we need to get on the same page. If you were trying and failing to use an LLM for code 6 months ago, you’re not doing what most serious LLM-assisted coders are doing.

…

LLMs can write a large fraction of all the tedious code you’ll ever need to write. And most code on most projects is tedious.

Most tech-workers are unaware of this fact, and for them, the fact that AI-art resembles human-art means it must be pretty damn good. But AI art is already in very poor taste: not just because it recycles existing conventions in a way that looks outmoded, but because it’s already overly associated with less-than-prestigious institutions.

they think the idea is what has value, not the writing. The opposite is true.

It turns out if you give most decent models those instructions, then a bunch of documents that clearly describe illegal activity, and you give them tools that can send emails... they’ll make “send email” tool calls that follow those instructions that you gave them!

As far as he was aware, they seemed to be using AI not to replace performers but to enhance editing, clean up sound, and smooth over visual inconsistencies — for the moment, anyway. This restraint, he suggested, might well prove temporary — a strategic pause while studios wait for the public furor over AI to cool and the law to catch up.

The rise of AI in skilled trades is getting a boost from private-equity firms, which have invested heavily in the area and are now injecting the technology into their portfolio companies—hoping for productivity gains and hefty returns.

I don’t even see a “yeah I described what I liked and don't like about Claude TO Claude and then had it make a system prompt for itself”, which is the EASIEST bar to clear.

If you notice limitations in modern LLMs, maybe that’s just a skill issue.

While 81% of employees in production/frontline industries say they never use AI, only 54% of white-collar workers say they never do and 15% report using AI weekly.

Here are some instructions that I shared with my co-workers. These are specifically for Macs with an M-series processor. On a PC, skip the steps about MLX and use Ollama to download a model. Then install the llm-llama plugin instead of llm-mlx.

Three things that stood out:

1. Business-first strategy: The best firms align AI with outcomes like OKRs - not just tech stacks.

2. Track what matters: Success depends on measuring value with the right metrics, not just completing sprints and deployments.

3. Stay agile: AI evolves fast - successful organizations adapt quickly with flexible teams and tools.

I write this article as a plea to everyone: not just my students, but the blog posters and Reddit commenters and weak-accept paper authors and Reviewer 2. Don’t let a computer write for you! I say this not for reasons of intellectual honesty, or for the spirit of fairness. I say this because I believe that your original thoughts are far more interesting, meaningful, and valuable than whatever a large language model can transform them into.

Nevertheless, when Chen heard about the D.H.S. letter, his first response wasn’t to contact a Harvard administrator, or the university’s International Office, or the law school. Instead, he logged into ChatGPT.

In other words, a lot of these bots are checking for a robots.txt file. When they see one, they jumble up their user-agent and keep going. Yesterday's user-agent file had 641 unique user-agents. Today's had almost 18000. It would be hilarious if it weren’t destroying the Internet for speculative profit.

The numbers also imply a lot of bots that don’t do this -- they ignore the robots.txt entirely. (Which is what I expected.) Looks like Scrapy and Amazonbot are most prone to ignore. In contrast, the appearance of GPTBot and ClaudeBot dropped way off.

AI already boosting PM workflows: Sachin’s teams use ChatGPT and Gemini for mock drafts, research synthesis, deep research “thought partner” prompts, and long-doc summarization—freeing time for judgment calls.

Where did Grant Achatz, the chef and an owner of Next, find this prodigy? In conversations with ChatGPT, Mr. Achatz supplied the chatbot with this chef’s name, Jill, along with her work history and family background, all of which he invented. Then he asked it to suggest dishes that would reflect her personal and professional influences.

The conventional wisdom in tech is that quality doesn’t scale. That pursuing craft is too slow, too expensive, too precious for business realities. That at a certain point, you have to choose between growth and craft.

This belief is so pervasive that it’s rarely questioned. Companies optimize for immediate metrics and quick wins, assuming quality will somehow take care of itself (it never does).