Category: Uncategorized

I think that this was this Computerphile video: https://youtube.com/watch?v=-0HRzXk8vlk

https://crikey.social/@ross/117034108905770798

I’m so stuck. I read an article this last… week? Fortnight? About LLM coding agents and how they created so much context. It had lots of numbers about how a 50-token short question adds up all the files and inputs as tokens of input.

Do you know where this is? It’s not in ANY of my digital brain!

I sometimes have a worry: What if I’m wrong about LLMs? What if it’s actually good and I’m missing something and get left out?

And then I see something like this and I feel very justified:
– Generative AI just bad and wrong a lot of the time
– We still have to do The Work. Making this map WAS The Work. If the person made the map by hand they would have learned where the countries are as they did so. The Work – however mundane – is still important.

https://www.theguardian.com/us-news/2026/jul/30/government-map-mislabels-african-countries

We should probably do something about this. 🤔

https://mastodon.energy/@CLEW/117013236245639511

The book doesn’t seem to have much reference to AI, but Cat’s (@grimalkina) “Brains in jars” model of software teams seems to describe exactly what LLMs might be (are?) good at replacing.

“Interactions, social learning, and shared problem solving are invisible in this model” – and seem to disappear when agentic coding takes over.

“good psychology is _how_ good software work happens.”

👀

Reading @grimalkina’s “The Psychology of Software Teams” (because I’m VERY interested in the people side of software development)

It looked like quite a dense, academic book. But it’s actually short, punchy, and very well written (so far!)

Today was one of those “Glad I know lots of weird stuff about computers that not a lot of people know” days.

Enjoying doing deep work with the deep knowledge. 😃

Oh gah! UK Plug-in Solar will be legal from 27th August but does NOT include kits with batteries. It’s panel+inverter only.

https://solarenergyconcepts.co.uk/post/plug-in-solar-uk/

I spent some of yesterday watching the Patreon guy talk passionately about taking risks, putting it on the line, and doing what’s hard.

https://youtu.be/17_HcR95YBc?t=3390

Today I read that he’s laying off 20% of his workers.

https://www.theverge.com/tech/970211/patreon-layoffs-ai

And now I’m very confused about this guy.

Finally watched Ready Player One last night. I’ve not read the book.

Was disappointed that it just seemed to be a blockbuster CGI fighty thing.

Is the book a more interesting story?

Thinking about this… it if I have included an open source license then they are allowed to create a derivative work.

And they are supposed to keep a copyright notice, so I must be giving consent for that personally identifying information to be used.

So this seems like a legitimate thing to do with open source code, doesn’t it?

Regardless: this stinks – not least because they have included code of mine that has no license and thus is NOT open source. 😡

https://mastodon.social/@concretedog/116976794778175500

We finally finished this epic name-grouping quiz today! Only took us 7 months and 1509 mistakes. we really weren’t very good. But I can close the tab now! Yay!!!

https://thomaswc.com/2025.html

I’m sorry, what now???

“Every time an EV connects to a charger, it establishes an IP network over Power Line Communication (PLC).”

A mini network between your car and charger! Fun!!

“This network was designed for V2G protocols. But we also discovered other EV charger operating system services listening on these network interfaces.”

🫣

https://www.saiflow.com/blog/the-hidden-ccs2-attack-surface-on-ev-chargers

https://social.jpoesen.com/@jpoesen/116956567980828772

I do wish @calebporzio was properly on these socials. The journey of self-discovery he went on in his last two podcasts was epic. I like where he landed. He is not Ian. He works differently. He finds joy in different things. That is a good thing. Stop trying to be someone you are not!

I hate to do this, but this is a test!!

I hear that this new movie “The Odyssey” is based on a book.

Do you think I should read the book before seeing it at the cinema?

🤔

I find myself wondering if there is a job in optimising LLM agent skills by converting them to run deterministically with real programming/scripting languages.

I see so much “if…then”, decision trees, and even literal “for each” mechanics in instructions to LLMs.

This week I reduced a part of a skill that took 14s to run with an LLM to a PHP script that took 2 seconds (mostly querying an API) and used zero tokens.

With companies asking people to reduce token use, this seems like a no-brainer.