Tyler Ham
tylerhamilton.net
Tyler Ham
@tylerhamilton.net
In the 2000s it was neat to read coverage by Ars Technica of studies/research about the productive and useful length of copyrights. IIRC it was 15-40 years. Unfortunately the rich holders (eg Disney) have the incentive to push that.
Why are we following the copyright law set by the rich companies?
January 27, 2026 at 10:25 PM
Our definitions of copyright and theft are coming from from laws: written and as observed & refined by courts & case law.

This is backwards and needs to evolve.

Laws are old and slow to change. Case law slightly less quick.

Not saying it should be a free-for all.
January 27, 2026 at 10:19 PM
Limited environments is level 4, Unlimited environments is level 5. For better or worse, SAE doesn't distinguish or define anywhere between the wide range of limited vs unlimited.
January 27, 2026 at 10:04 PM
Is this different from an existing driverless robotaxi?
January 27, 2026 at 7:06 PM
"Books" is a pattern of a barrier to entry for the quality of the content. I have found a similar thing looking for reviews of products: magazines and YouTube videos are high barriers to entry and those don't have as much slop or SEO as review websites.
January 24, 2026 at 5:17 PM
Good recipes are likely in books.
The creative, incentive, and innovative recipes will hopefully stick around, and be first in forums/Reddit and later on passion project blogs that don't need or want ads or SEO.
Later they can be in books too.
January 24, 2026 at 5:14 PM
Google made it even more obvious when you click over to "AI mode" it's noticeably better if slightly slower.
January 24, 2026 at 2:16 AM
I think this was my consultant-speak coming out 😂
"It depends" was another thing I learned early as a consultant.
January 21, 2026 at 9:54 PM
Fair! To be clear, I don't believe Microsoft when they sell .NET or Azure. I read a variety of sources and make a pragmatic decision. Even so far as to possibly go Polycloud.
Likewise I don't believe Ford or Tesla but I do enjoy reading Consumer Reports or Car & Driver. Etc.
January 21, 2026 at 9:51 PM
But it is worth evaluating how AI assisted coding may help speed upskilling, training, debugging and understanding the specialties. Each business can make the ongoing evaluation as the tools and business changes.
January 21, 2026 at 8:40 PM
Fair, but what is your plan for when a specialist gets hit by a bus?
Is it so specialized it takes months to hire and train if someone quits?

In this case, perhaps the slow down tradeoff is in the long term interest of the business. Nobody is claiming these tools can do absolutely everything.
January 21, 2026 at 8:37 PM
Globally distributed teams have similar considerations: when my coworkers in India write code, I will be working with it and adding to it hours later while they sleep. And vice versa.
January 21, 2026 at 6:51 PM
I'd love to see an estimate of how many hours of work a coding agent saves, then it would be easy to calculate calories and then energy & water savings from a standard diet.

Ofc, you're not actually saving energy but it'd be like taking a car a long distance vs walking the same place.
January 21, 2026 at 5:00 PM
I'm aware of the appeal to authority fallacy but as I look for writing and discussion from authors, people who have written languages and frameworks, etc, I have found the balance to be in favour of LLMs for coding. Not universal and some caution but overall more for than against.
January 21, 2026 at 8:56 AM
There just ends up being a lot of code to do all these simple things. It's not very difficult there's just a lot of the same repetitive types of coding:
- move data center to cloud
- backend API for a smart oven
- code to set up AWS infra
- zip up music files
There's just a lot of that in the world
January 21, 2026 at 8:52 AM
I think this may be one of the reasons why we're disagreeing: I totally can see how ML and academic research is novel code that's not well written by LLMs.
Most of the code that I've seen is almost boilerplate and not overly complicated:
Move this data here to there. Make a website to do this.
January 21, 2026 at 8:48 AM
This is true! It's like hands free cruise control available in many new cars. The person in control always needs to be 100% responsible for all actions and may be lulled into distraction, which is the most dangerous time.
January 21, 2026 at 1:21 AM
Maybe! But at this point neither of us have shared our professional experiences and you've dodged the question so I'm left to wonder if you are making statements about something you have not used the latest versions of as designed for professional SWEs. Please correct me if I'm wrong.
January 20, 2026 at 7:45 PM
How many hours have you used a recent (past 2 months: Gemini 3, Opus 4.5, Codex 5) tool to deliver software?

Eg I used to dislike Java then I came back to it and it had improved (eg generics) and I realized it wasn't so bad when used in the right way.
January 20, 2026 at 7:16 PM
Also, Codex 5 and even more Opus 4.5 are actually very notably better and weren't available 5 months ago.

Eg, It is now possible to quickly automate small daily tasks we always knew were possible but never worth doing.
January 20, 2026 at 7:09 PM
I 😂 because @t3.gg , who has typical YouTube video titles, is one of the reasons how I increasingly became interested in using the tools more and more.
To be clear, this is after I started using some of the tooling in depth myself.
January 20, 2026 at 7:06 PM
Yes, never said he wasn't. Most SWEs I talk to and read the writings of are taking this approach. However using it a bit shows us that it's pretty good and over months we use it more and more, especially at the rapid rate LLMs have improved over the past 6-12 months.
January 20, 2026 at 4:03 PM
Welp, most of the Internet runs on Linux. Guess we can see how it'll play out.

Even the Bluesky devs are using LLMs now, you can see them talk about it on here.
January 20, 2026 at 8:31 AM