AI for the rest of us

Welcome to the AI community for everyone.

Hello friends,

This week we held our September meetup in London where we discussed how AI is changing recruitment. The most surprising thing about the discussions that followed was that there was real consensus, not disagreement. AI didn’t create many of these problems, it just made them faster and larger. Sloppy CVs have always been a thing and CV review has always been a flawed process. The advice wasn’t particularly new either - relationships, reputation and integrity are as important as ever.

If you’re job hunting at the moment it’s time to invest in your network. Get out there and meet people. Ask for that referral. Understand the unique talents that will make you stand out… and here’s the awkward part… tell people about them so they can advocate for you.

Charles has been in Copenhagen this week for GOTO. Amongst other things he had the pleasure of doing something which felt delightfully meta: an improvised unscripted interview with software industry legend Kent Beck about improvisation in music and what it teaches us about programming. It was also fantastic to catch up with friends and colleagues from across the industry.

Have a wonderful week!

Hannah and Charles

The Hard Part of Agents is Integration

Anyone can build an agent. Very few can make it work inside years of accumulated business software. That's where Focused does its best work.

Contact Focused

What’s Hannah reading this week?

It’s going to be a very busy couple of months and I’m pushing myself to achieve a lot by the end of the year. This week I decided I won’t get anywhere quietly grinding away, alone at my desk. I’ve got a much better chance of success if I tell people what I’m working on and ask for the help I need. If you follow me on LinkedIn you’ll see a lot more from me, because I’ve committed to posting every day for the next 8 weeks.

Follow along while I try to achieve the impossible and have a bloody good time doing it.

The AI news cycle this week has been dominated by the chaotic thrashing in the US. On Tuesday we saw Trump embrace leaders from OpenAI, Anthropic, Nvidia, SpaceX, Meta and Google with the meeting concluding with a short co-signed statement around AI safety.

The gist of the agreement was that the companies are responsible for ensuring the safety of their own AI technology. Self-regulated rather than state regulated. They committed to work with independent auditors and create a review board, all good intentions. It remains to be seen if these good intentions will manifest in practical changes and better oversight. Without enforcement it feels a bit weak: "Just don't do anything bad, ok?"

If you want to read the full document (it’s only one page) it was posted on X by the White House. I honestly couldn't find a more official or authoritative source and I refuse to share a link to truth social!!

Then, just one day after the White House meeting the US FTC (Federal Trade Commission) launches an investigation into whether the AI companies are harming US consumers. Are they friends? Enemies? Frenemies? Is it all just good old PR and perception management?

I worry that it’s all a bit US focused, with a single government taking an authoritative position on AI safety without global collaboration. Here in the UK the conversation about AI sovereignty has been provided with more fuel as global AI leaders cosy up with Donald Trump. Optionality between AI providers feels like a sound strategic move, not a future concern.

Oh and we’re all supposed to call it Super Intelligence now. Sigh.

The narratives we’re hearing around AI right now feel deeply unhelpful. Whilst loudly beating the drum for safety and regulation Anthropic have also been lobbying religious leaders to consider AI consciousness. WTAF. Just because a computer program can mimic some traits of our human experience does not make it more than a computer program.

Our perception of these computer programs matters right now and the anthropomorphising of bits and bytes in the general public prevents sensible dialogue. The AI giants and the media are both to blame here, with their incentives misaligned with the needs of the average user or reader. Sensational stories get clicks and we all need to remember that.

On a more positive note, I read a lovely short article this week about spec driven development in the real world. Chad Fowler CTO of Blueyard writes “The Spec Can Come Later”.

“Experienced developers know this in their bones. Programming is one of the most important ways we find out what we actually want. You try something, it's wrong, and you change your mind. Some requirements don't exist until an implementation breaks them.

Here's a rule I know to be true: My first attempts at specifying a system are always wrong. This is a fact I've treated as a rule since I first started programming in earnest in the 90s.”

Build, measure, learn. That’s what we’ve been saying for years. The tools and practices change but the learning loop remains. Figuring out what we need to build is the hardest part of building. As Charles shares below, the speed of coding is not translating into business outcomes, perhaps the speed of learning is now often the limiting factor?

What's Charles reading this week?

I’ve not had much time for either reading or writing this week, but I have had an intense few days of listening, talking and learning at GOTO Copenhagen.

Of course conferences are their own bubble and this experience won’t be universal, but the rate of change and the speed at which the IT industry is trying to adapt to it is truly astonishing.

During the conference we heard examples of dark factories, where AI agents build software and check each other’s work with minimal human involvement, as well as examples of keeping the humans involved throughout, applying good engineering practices from the last 40 years. A lot depends on the context. If the system you are building is critical, then a high level of human oversight is likely needed. Conversely, if you are constrained on human review and working on a less critical system, then it might well be that sacrificing that review in order to go faster is the better option.

Of course, if you aren’t bottlenecked on producing code then producing code faster isn’t going to help. AI coding is allowing individual developers to build much more ambitious applications than before, particularly hobby projects, and write in languages they don’t really know. Equally though a message that came up in multiple talks I went to was that whilst individual developers feel more productive, organisations aren't (yet) seeing the gains.

Anecdotally from speaking to attendees, I got the sense of a rough 50/50 split between those who were excited to be working with the new tools and those who felt their jobs had become much less enjoyable. I think that for those readers who manage software teams there’s a lot to think about in terms of supporting engineers through this extraordinary, rapid change. As AI permeates other professional disciplines, that same feeling will spread further.

Alongside the AI talks was a whole mini-track around ethics, sustainability and responsible tech. I was thrilled to have a full room for my own talk on sustainable AI, and loved hearing from Wilco Burggraaf on how he and his team used AI agents to find performance and energy savings in PyTorch - it’s brilliant work.

Rebecca Parsons, CTO Emerita at ThoughtWorks, made the point that building responsible software takes active, ongoing consideration of values, unintended consequences, and negative impacts on all stakeholders, not just good intentions. The problem is usually that people either don't think about it, think it's too expensive, or make trade-offs without seeing the full picture. She also made the point that in a global world differing cultural norms make understanding where the lines need to be drawn challenging, since cultures have very different ideas about what is and isn’t acceptable. An overall message here was that technologists should hold themselves accountable. Our job is to make clients aware of the risks (she described a team who kept questioning a client until it dropped a bad idea). We should be accountable for both the good and bad outcomes of what we build.

On the final day Tim Berglund, who gave a fun and entertaining talk on the history of AI, suggested it will take 5-10 years for the new practices to settle out but he views this as an extremely exciting time, and suggests we should all be building, experimenting, and learning together. I’ll add that I think consciously maintaining human connection alongside this feels more important than ever.

Kent Beck closed out the conference. His main messages were about setting boundaries with the genie - his term for AI. Beck explained that he is bipolar, and has found programming with AI has led to manic episodes. His point was that the decisions the tool makers are making have real consequences and that the people who are building them have responsibility for that. Beck wants new AI tools that promote connection, automate toil, preserve struggle, surface choices and encourage learning.

Updates

Early Bird Tickets

Our next 2 day conference is set for 25-26 February 2027. Grab your early bird ticket for £150.

Get Your Ticket

Between Jobs?

Interested in a more human approach to career change and AI? Join the Surviving The Gap community for practical workshops, useful resources and early access to Career Core.

Learn More