Latest blog posts
Maximal Instruction Prompting: a strategy for software development with LLMs
I've been trying new ideas to see how we might innovate software development using Large Language Models (LLMs). This article introduces a new concept, Maximal Instruction Prompting (MIP), as a strategy for software development with LLMs.
I've used this approach for several projects, and the concept continues to evolve with each new use. However, even in these early stages, I've seen great results.
This article describes the MIP strategy and the benefits it can offer. It provides more detailed explanations and ends with an example of an application built this way.
Can my tools build tools? Pre-rendering web pages with help from ChatGPT
We're all trying to work out the implications of generative AI. We've all seen examples of it building websites and coding a version of the snake game. These seem impressive until we realize there are millions of websites and thousands of versions of snake, so it's not too surprising tools like ChatGPT can regurgitate them. After all, it's easy to replicate something that's well-understood?
How about a more interesting challenge? How well can these tools build something new? I decided to give this a try. I needed a tool to use with my blog site, davehudson.io. How well would ChatGPT be able to help?
Understanding other people's code
A few days ago I decided to move away from my old Joomla-based blog site and set up my own static web site. After some investigation I settled on using Hugo. I was impressed by some of the web sites I'd found that were using it. Also my expertise in CSS, and JavaScript were pretty limited and Hugo didn't appear to need much of either.
As I started to build the site I found myself treading a famililar path. Every example I looked at did things in different ways so it was hard to work out how to put the learnings together in a coherent way. As is the norm for busy engineers, the developers got things to work and then moved on to other things. "Documentation is boring!"
In some instances I could look at git histories and discern more of the developers' intents, but all too often I ended up with a frustrating effort to reverse engineer what they'd ended up building. It's not like this was a surprise - I've done this dozens, maybe hundreds of times. But I wish, as an industry, we could do better. I've worked with many engineers and tried to persuade them we should, so maybe I can persuade you too?
What IoT history reveals about blockchain's challenges
2009 saw Satoshi Nakamoto deploy the first Bitcoin node, and within five years its blockchain had become a large-scale industry.
Early enthusiasm for new technologies is nothing new. With most, an initial wave of excitement sees new ideas touted as solutions to a huge range of problems, the hype fades, gives way to skepticism, and ultimately, real applications.
In the late 1990s, the idea of Internet-connecting every electronic device seemed inescapable.
Every vending machine, coffee pot, toaster, refrigerator, microwave, and TV, would be cabled to the “net”, and a utopian sharing of data would improve life for everyone.
The reality for what we now term the “Internet of Things”, or IoT, was a little different.
A market for Bitcoin transaction fees?
Most participants in the ongoing Bitcoin block size debates have a point of agreement; that a shortage of block space should have an effect on transaction fees. Arguments aside, then, let's see what has actually been happening. Are fees going through the roof? Are miners going to be celebrating a potential offset to the block reward halving that looms in July 2016? The results seem a little surprising!
More blog posts
You can find older blog posts on this page: Blog