ColinEberhardt

Oh wow, I loved AMOS - it is what got me seriously into programming in the first place.

Very cool. By the way, you can render many more datapoints on mobile if you use WebGL. Here’s a similar example - embeddings rendered using a T-SNE layout

https://blog.scottlogic.com/2021/10/15/efficiently-loading-m...

I created my own static site generator about 10 years ago. It was only around 100 lines of code:

https://github.com/ColinEberhardt/tiny-ssg

I used it for a few websites, viable replacement for Assemble and other SSGs of that time.

I presume it only recognised the BBC journalists efforts as satire due to the article in which he clearly states that this was his intention? Without that, I’m am confident it would have fallen for it.

Agents Skills are built upon “Five design decisions [that] are the load-bearing ones”

And Open Design (HN front page yesterday) is supported by “Six load-bearing ideas”

The similarities in the way these prompt libraries are documented doesn’t feel coincidental.

Funny coincidence - I visited that valley for the first time just last week, on my way to the Isle of Skye.

And as a further coincidence, I met Jimmy Saville about 25 years ago. I was in Leeds hospital after a heart operation, and this old and somewhat scruffy track suited guy just walks in to the ward and starts talking to me. I had no idea who he was. After he left, a nurse asked “did you speak to Jimmy?”. It was creepy and unnerving seeing first hand how he just got to roam around.

I can confirm, the graffiti-covered Saville residence has almost completely been demolished.

Agreed, I could just about bear it until I hit the “ Six load-bearing ideas” section. Very off-putting.

Very true, with the right feedback loop AI would do a wonderful job of refactoring.

But if AI is the primary author and consumer of this code, that would be an unnecessary step. No need to clean it up for our feeble little human minds.

I was just interested in what this file actually does - and am finding it hard to grok, scrolling through on a mobile device!

Can someone please use AI to explain this code smell?

I used to find Gary Marcus a good antidote to the AI hype, and followed his critique. But honestly, his more recent writings are clutching at straws. This article feels like desperation.

It’s a bit like saying that driving cars still requires human muscles to operate the controls, so human strength has ‘won’, when it is clearly the internal combustion engine that has created the speed advantage of the car.

Looks interesting. Quick question - one of the biggest challenges with agentic systems in non-deterministic behaviour. Does this framework do anything to address this? Does it help test and validate agent behaviour?

Bloody hell - that hits hard

I agree, timelines are fascinating. I did my own research and built a simple visualisation of the changes in time zones over a 120 year period:

https://blog.scottlogic.com/2021/09/14/120-years-timezone.ht...

One drunken night in the company of Norman Lovett was enough for me :-)

My random claim to fame; I was the support act (juggler) for Norman Lovett (the red dwarf ships computer), for one night only in the Welsh town of Bangor.

What a life I’ve lived.

The same minion that came up with the cute effect that covers your screen with the word DEVELOPERS, when you scroll to the end of an article?

Bingo! Loved that show

Wow, that's an impressive amount of dedication, but I guess you need that if you want to set a world record

Dedication, mmm, dedication. Dedication, that’s what you need. If you want to be the best, and if you want to beat the rest. Dedication way you need.

Hopefully that means something to Brits of a certain age ;-)

I know it’s a minor point, but it bugs me every time this form pops up…

Captive (noun): a person or animal whose ability to move or act freely is limited by being kept in a space; a prisoner, especially a person held by the enemy during a war.

Not an ideal term to use from a user perspective.

Nice work!

D3fc maintainer here. A few years back we added WebGL support to D3fc (a component library for people building their own charts with D3), allowing it to render 1m+ datapoints:

https://blog.scottlogic.com/2020/05/01/rendering-one-million...

An important point here is that it isnt doing a 1-shot implementation, it is iteratively solving a problem over multiple iterations, with a closed feedback loop.

Create the right agentic feedback loop and a reasoning model can perform far better through iteration than its first 1-shot attempt.

This is very human. How much code can you reliable write without any feedback? Very little. We iterate, guided by feedback (compiler, linter, executing and exploring)

Looking at the commit log, almost every commit references an AI model:

https://github.com/nibzard/awesome-agentic-patterns/commits/...

Unfortunately it isn’t possible to detect whether AI was being used in an assistive fashion, or whether it was the primary author.

Regardless, a skim read of the content reveals it to be quite sloppy!

Ironic, yet practical. This is so that the issue can be 'pinned' to the top of the Issues page, making it more visible:

https://github.com/ghostty-org/ghostty/issues

Nolan shares some interesting reflections at the end of this post. Here are a few direct quotes:

"I’m somewhat horrified by how easily this tool can reproduce what took me 20-odd years to learn"

"I’ve decided that my role is not to try to resist the overwhelming onslaught of this technology, but instead to just witness and document how it’s shaking up my worldview and my corner of the industry"

"I have no idea what coding will look like next year (2026), but I know how my wife will be planning our next vacation."

It’s still around, and tends to be adopted by big enterprises. It’s generally a decent product, but is facing a lot of equally powerful competition and is very expensive.

“When an LLM can generate a working high-quality implementation in a single try, that is called one-shotting. This is the most efficient form of LLM programming.”

This is a good article, but misses one of the most important advances this year - the agentic loop.

There are always going to be limits to how much code a model can one-shot. Give it the ability to verify its changes and iterate, massively increase its ability to write sizeable chunks of verified and working code.

Oh, so _that_ is what a sub-agent is. I have been wondering about that for a while now!

Likewise. I have a nasty feeling that most AI agent deployments happen with nothing more than some cursory manual testing. Going with the ‘vibes’ (to coin an over used term in the industry).

We find testing and evals to be the hardest problem here …

I wonder what this means for the agents that people are deploying into production? Are they tested at all? Or just manual ad-hoc testing?

Sounds risky!

Neat! Really like this, and will be interesting to see how it tracks over time.

My only concern is the scale you have selected as we are already close to 100, are you going to have to allow it to be ‘turned up to 11’ at some point?