AI for the rest of us

Welcome to the AI community for everyone.

Hello friends,

Hannah was sitting in a cosy cafe, gazing out at a rainy fjord in Norway when last Sunday’s newsletter hit your inboxes. She only realised then when a flurry of ticket sales for AI for the rest of us, London 2027 caused her watch to start buzzing with notifications! What an incredible vote of confidence - thank you to all the readers who grabbed their tickets!

If you missed the community tickets last week then there are 4 left at £96 each.

Don’t forget that we’re also getting together on Tuesday evening for our next evening meetup. We’ll be talking about the way AI is changing the dynamics for job applicants and hiring managers. Did AI break recruitment? Come along and share your thoughts.

Charles has had a fairly social week, in between desperately trying to get ready for conference season, starting with GOTO Copenhagen next week. He went for an excellent curry on Brick Lane with work colleagues on Wednesday, had lunch with a friend on Thursday, and was in the Twofish studio on Friday working on their much delayed third album. His eldest is back up at university starting second year. It's been lovely having her home over the summer, and the house feels very different without her. She seems to be thoroughly enjoying herself though which is fantastic.

Have a wonderful week!

Hannah and Charles

Community Tickets

Grab an exclusive discounted ticket for AI for the rest of us 2027 on Feb 25-26.

Grab your ticket now

What’s Hannah reading this week?

This past week has felt more like a month. I’ve been sweltering in saunas, leaping into fjords, gazing at waterfalls, wizzing around on speed boats, climbing mountains, riding trains and enjoying an excessive amount of cinnamon buns. Norway did not disappoint.

On Saturday we checked into an impossibly beautiful little red house on the Aurlandsfjord and were greeted by a rainbow stretched across the water between the valley sides. A sign that we were about to start an unforgettable week. The place is like something from a fairytale, I’ve never seen so many waterfalls. Everywhere you look there’s another cascade of water falling from hundreds of meters above you, bouncing down the mountainside into the deep fjord below. At night I slept with the window open so I could hear their distant roar as I fell to sleep.

With the tech industry forecasting an AI apocalypse I visited a place that did not give a crap about Artificial Intelligence. Will AI touch the little red house? Will it touch the tiny village only accessible by boat? Will it change the landscape we walked?

I’m not naive. I know that supply chains are a thing, and that recessions impact tourism, which impacts the communities that rely on tourists. But, I spent a week surrounded by people who won’t be left jobless by AI. They are not worried. They are living beautiful lives in a beautiful place and I think I may do the same if AI steals my job.

This trip to Norway marks the end of my summer frollicking and the beginning of a whirlwind conference season. I have 9 speaking gigs in the next 8 weeks taking me to London, Belfast, Edinburgh, Malmo and Salt Lake City. At home I’m referring to this as the BIMP debut tour because I’ll be representing my own product for the first time at all of these events. It’s very exciting!

As I’m sat at the kitchen bar writing this newsletter my partner is setting me up with an AI assistant to help me stay on top of things while I’m on the road. I’ll share how that’s going in future newsletters.

With very little time for reading this past week I have just two tibbits for you this week. The first is from product guru Melissa Perry, author of Escaping The Build Trap. She reflects that the Build Trap, the way many organisations measure success in terms of features shipped, is getting worse, not better with AI.

“The question we kept coming back to: are we just recreating the Build Trap at a much higher speed?

Building has gotten dramatically faster. Teams can spin up features, prototypes, and whole products in a fraction of the time it used to take. But I don't see most companies learning any faster” shared Melissa on LinkedIn.

More is not necessarily better and strong product leadership must have a deep understanding of the user, regardless of the domain. We joke at home that BIMP was the product that Stuart and I were inevitably going to build - one product leader from the world of Kubernetes and Infrastructure and the other from the world of Platform Engineering, Container Security and Vulnerability Management. We know these users better than anyone and we can build for their needs. No AI tool can replace that customer empathy!

The second conversation that passed by my feed was the use of formal verification methods, Lean and TLA+ (Temporal Logic of Actions) to validate and uncover issues with AI architected solutions. I had never heard of TLA+ before but apparently Claude is excellent at writing it. Boris Cherny, Creator and Head of Claude Code shared the outcomes of his experiments on X this week which has created a swell of interest with other software developers giving these pre-AI methodologies as spin with promising results.

What's Charles reading this week?

I’ve found myself this week reflecting quite deeply on a job I’ve been doing for a client for most of this year. I’ll spare you the details but the gist of my role has been to act as a sort of translator, taking complex thinking and research around generative AI and then trying to describe and distil it in ways that make sense to different audiences such as program directors or policy makers. It’s sort of comms, sort of technical writing, and sort of journalism, and also a bit of business analysis, but it isn't really quite any of those things. It's work I love doing. Unfortunately for me though, it's very hard to explain either what it is or why it can be so valuable, which makes it hard for me to land these sorts of jobs. (Side note - if you’d like me to do this for your project, get in touch!)

Given this I’ve been thinking about writing a blog about this role and why an LLM can’t do it, starting with a quote from American computer scientist and mathematician Leslie Lamport, who, riffing off one of my writing heroes William Zinsser said, “Writing is thinking. If you aren’t writing you only think you’re thinking”. I will try to write that blog at some point, but part of the argument I wanted to make was that LLMs are weirdly bad at writing. The software industry is however fortunate to have a number of very brilliant thinkers and communicators and, the hive mind being what it is, not one, but two of the best wrote excellent blogs covering that topic. So, whilst I might try to expand on my own thoughts at some point, for now I thought I’d just share theirs.

The first is from Charity Majors, co-founder and CTO of Honeycomb.io. In “Confessions of an Unrepentant Slop Snob” she offers a peek into the research behind Honeycomb’s “AI Norms and Values” docs. If you are grappling with AI use in your company, then this blog and the linked documents are a superb place to start.

There’s also a section in Majors’ blog where she talks about LLM generated writing and why she hates it:

“Language evolved as a way to connect — person to person, mind to mind, one mind to many. This is some of the oldest and strangest wiring we have as human beings, and the neurological infrastructure for linguistics gets used and reused, over and over.

n software, for example, we convert natural language into bits and bytes that computers can use to do math on. Lawyers convert language into legal text and taxonomies. The technical and professional worlds are awash in dialects where language has been abstracted from its roots as an emotional and relational tool and given functional, depersonalized meanings.

n many of these contexts, substituting AI-generated language can be wholly acceptable. No one blinks an eye if you use structured data generated by AI, or a formal proof generated by AI (as long as it’s accurate). The situations where the use of AI tends to land jarringly, causing frustration, rage, even a sense of betrayal, are the ones where the value of the communication is less abstract, and more personal or relational.”

As a sidenote, I find it almost miraculous that we can read the thoughts of people, and indeed listen to music, written by people who lived thousands of years ago.

Co-founder and CTO of Oxide Computer Company Bryan Cantrill’s contribution is a blog called "The revolt of the reader"

“I am writing this now because — speaking as a reader — we are exasperated: too many people — people we otherwise like and respect! — are writing (or otherwise putting their name to) pieces that are clearly LLM-authored. We readers are left with pointed questions for those promoting LLM-authored pieces: do you think readers can’t tell? Or do you think readers don’t care?”

I confess I have tried using LLMs to write and found it unbearable, and as far as possible I avoid taking on writing jobs that require me to use an LLM to generate text. Confirmation bias is a thing here, but my own experience is:

My brain shuts down the moment it detects LLM-generated writing. If I’m asked to edit generated writing I have to force myself to do it, and it takes so much more effort and so much longer for me to do than hand-written human prose: no matter how poor the latter is. Given this, I would much, much rather be sent your poorly explained thoughts to edit than ones you got AI to “polish”.
I also do not find using an LLM to generate text helps me from a productivity point of view. Time to first draft is obviously faster, but it then takes me so long to get it to second draft it actually costs time.
No matter how hard I try, I don’t think the results of an edited LLM-generated article are ever as good as if I just do it by hand.
Given that roughly half my income is made through writing, not using an LLM to do it was a tough line to hold last year since most writing gigs wanted me to use AI. This year the pendulum appears to have swung back at least part way. There’s hope for me yet!

As another side note, the number of AI-generated slop messages I get on LinkedIn these days is just ridiculous: “Hey Charles, your focus on green software caught my attention…” Aaaah! Equally though, it will never be cheaper to learn how to use AI in your work, so experiment as much as you can and find where the value is (and isn’t) for you.

Somewhat on the writing theme, Google and Google DeepMind researchers launched the DeepMind Institute on Wednesday to advance the conversation around artificial general intelligence (AGI). There’s no good definition of AGI as such, but it’s best thought of as the point at which an AI system can understand, learn, and perform any intellectual task a human can. DeepMind chair Demis Hassabis believes we are close to this.

The institute lists DeepMind co-founder Shane Legg, Google executive James Manyika, and Hassabis as directors, with Legg serving as managing editor. It aims to surface differing views between Google, Google DeepMind, and the broader global research community around AGI. “They will not always agree, and they will likely change their minds, as more data and information comes to light at the fast-moving frontier,” the announcement read.

Along with the announcement are six essays. One, by DeepMind safety researchers Rohin Shah and Anca Dragan, argues that AI’s shrinking window of transparency — the ability to see and check a model’s step-by-step reasoning — is not inevitable. Another makes an economic case which is broadly designed to minimise social backlash and unrest from the AI transition by gradually easing society into universal basic income. I don’t agree with everything here, but there’s plenty of excellent food for thought in the initial set of articles.

It's been a tough year to be a climate change optimist and work in software. So whilst about as effective as playing a Kazoo in a thunderstorm (i.e. not very) I nevertheless gave a small cheer for New Jersey, who ordered the operator of one of the US East Coast’s largest planned data centres to pay a $1.1 million fine for secretly installing and operating gas generators in violation of the state’s Air Pollution Control Act. DataOne got hit with the fine after an investigation by The Guardian and Floodlight News in August shared thermal drone footage showing that 45 of the 62 gas generators were operating. None of the generators had the necessary permits, despite running at 1,982-kw capacities, i.e. more than 50 times higher than the state limit of 37-kilowatts.

Ars Technica’s senior policy reporter Ashley Belanger quotes Nichole Gardner, a co-chair of a local environmental non-profit Sustain SJ, as saying: “We believe it is insufficient, both in dollar amount and enforcement. The DataOne site should be made to discontinue operations during the 45-day period until the proper permits are issued.” Belanger also drily notes, “In a statement to The New York Times, the company seemed unfazed by the fine”.

Updates

Join us on 29th Sept

"Did AI Break Recruitment?" - We discuss at our next community event

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