AI for the rest of us

Welcome to the AI community for everyone.

Hello friends,

This week we hosted a virtual meetup where we were joined by Emma Burrows, Founder of Rezonant, for a glimpse into the future of product management. Emma shared how she's building a product brain using AI. If you missed the meetup you can catch up with the recording on YouTube. Thanks to everyone who joined and asked such excellent questions!

Hannah has been on an emotional roller coaster this week. Aside from a summer cold which has been incredibly unpleasant, she’s been getting BIMP ready for beta testing… The invitations were ready to send to our friends, advisors and wait list, we had one more build to push to production and then GitHub Actions went down.

It went down for 7 hours and 26 minutes.

HugOps to the team managing the incident at GitHub. It’s a stressful job when things go wrong. But OMG what timing?! It could not have been worse for us. Hannah was ready to punch the wall so instead put her boots on and went out for a walk… then did laundry… then gave up hope and went to bed.

Thankfully things had recovered on Friday morning and we are back on track! If you’re building applications with containers we’d love for you to take BIMP for a spin, contact Hannah or join the waitlist and a beta invite shall be yours!

Charles went to the Royal Academy Summer Show last Sunday with a very old friend of his. If you’ve never been, basically the RA crams as much art into a space as possible, some of it great, some of it objectively terrible. Amateur artists jostle for attention alongside famous ones like Tracey Emin, along with celebrities like Harry Hill (rather good) and Joe Lycett (terrible). Given the density, you notice what your eye is naturally drawn to. Almost all the art is for sale, and if he had the money there are certainly some pieces he would have taken home, and some pieces so awful the two of them collapsed into hopeless giggles. There’s also enough weirdness for the experience to be entertaining in its own right. It’s on until 23rd August.

Have a wonderful week!

Hannah & Charles

Turn your scattered context into a self-driving product

Rezonant builds the self-organising context layer for teams and their agents. They're offering a free 30-day trial for select design partners.

Get Access

What’s Hannah reading this week?

I had a fantastic dream this week. It’s always annoying when people tell you about their dreams, so skip ahead if you’re here for serious AI content. It was proper weird and yes it does involve AI.

We’re living in the future. BIMP is wildly successful (of course) and we’ve grown the company into a secure global network of data centres, responsible for all software supply chains. We are the ONLY place you can build trusted software. But our data centres are under siege from relentless physical invasions by hacker robots. They leap like fleas, climb stairs and navigate rocky terrain, break in, steal and sabotage.

The only thing that can stop the hacker robots - Elephants.

Yes, Elephants are the only thing that the robots are afraid of. Perhaps their training data sees them as a protected species, or they’re just way too big to fight. Whatever the reason, we have had to employ Elephants to patrol the grounds around our data centres.

There are not enough Elephants to go around all the tech companies and so part of my job is to negotiate the relocation of some Elephants as part of a merger with another tech company. We want their products, but we also really need their Elephants.

I was managing an Elephant relocation as part of my cyber-security risk management.

I enjoyed this bonkers dream more than most Netflix dramas I watch. Anyway, back to AI news.

Recently I’ve been researching the changing shape of software development teams as the world adjusts to life with AI Coding Agents. This week I’ve seen two new bold experiments in the real world that I wanted to share with you.

At Camunda, a process automation platform, they’ve been advertising for a few roles. I was intrigued by the role “Director - Product Building” - That’s a new one!

“We are introducing a new role: the Builder, one holistic role that combines what used to be split across Product Management, Design, and Engineering, and uses an AI-first way of working to discover user problems, define solutions, build, ship, and maintain them. Builders work in small pods (~3 builders per pod), shipping in days rather than months.”

That’s a lot of roles, responsibilities, skills and domains to squish into one role. I can see why the line between product and engineering needs to move, I also believe that engineers with product skills will be invaluable. But, I wonder if these Product Builders are supported by a Product Director who is looking at the long term product strategy, if that will fall to this new director of Product Building?

I was very surprised to see design consolidated into these new Product Builder roles. If “design” in this context means uncovering user needs, figuring out how a feature is supposed to work, testing that hypothesis and then delivering the desired user experience, then yes I can see why design is within the scope of this role. But, to be successful I think the Product Builder would need to be working within a mature design system, where details like brand, layout, UI components etc are provided.

It’s a bold experiment and I love to see bold experiments. I’m quite sure that the “old” way of building product engineering teams is gone forever and we need to experiment our way towards the new normal. I commend Camunda for going all in with something different. If you know anyone at Camunda, I’d be very interested to chat to them about how it’s going!

The second experimental team is Uber. They’ve also coined the term “pods” to describe a small autonomous delivery team of two - one engineer, one business domain expert. Praveen Neppalli Naga, CTO describes the experiment on LinkedIn.

"The idea is simple. We handpicked ~30 of our most AI-proficient engineers (people with deep knowledge of Uber's systems) and paired each of them with a domain expert from a business function. Then we gave every pod just two weeks."

I’ve talked about the opportunity that Agentic AI has for business transformation, eliminating business non-value add (BNVA as it’s called by the lean six sigma crew) overhead. Think of BNVA as the necessary backoffice work that keeps a company running, but never directly impacts customers. Wouldn’t it be great if instead of busy work we were able to refocus people on activities that deliver actual customer value!

The amount of paper pushing and repetitive process orchestration that happens in the back office of large companies is insane. The opportunity to automate with agents is a no brainer but it needs engineering time, and engineering tends to be focussed on product not operations. Perhaps it’s time to change that balance. Yes, build a successful product, but while you do that you should also build a successful company. Those are not the same thing.

Speaking of building agentic businesses, Y Combinator have open sourced their agent platform QM calling it a “multiplayer agent harness for work”. The best people I know, and some of the leading voices online, seem to have figured out the same thing. We deliver as a team. Having an agent as a personal assistant only gets you so far, using agents to eliminate handoffs, managing teams of agents as a team, and building systems around those teams of agents is going to have a far greater impact than local optimisation of each individual.

People have been commenting on the incredible salaries offered by Anthropic, especially after the comment made by CEO Dario Amodei, that people might be taking the jobs for the money and not the mission. No shit! The salaries sounds like a lot, but having worked for a few US tech companies I wasn’t at all surprised seeing packages over $500k for senior engineers. What did surprise me this week was the £44,620 - £49,065 offered to the Delivery Advisory role for the AI Safety Institute’s (AISI) Red Team. One of the most important AI safety teams in the world. I read the job description and felt unqualified for the work, and yet I’ve paid interns more than this in my previous roles. How can the public sector ever compete for talent if these are the salary bands?

What's Charles reading this week?

Google has announced some major leadership changes at DeepMind. Demis Hassabis, who has led the British AI lab since before Google acquired it in 2014, has stepped down as CEO to become chairman of Google DeepMind and chief scientist of Alphabet. Hassabis will also continue to lead Isomorphic Labs, DeepMind’s drug discovery spinoff. Koray Kavukcuoglu, who had been serving as Google DeepMind’s chief technology officer and Google’s Chief AI architect, will lead Google DeepMind as senior vice president.

Simultaneously, Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le, some of the most senior and high profile engineering and research talent in Google, have left to co-found a new autoresearch startup Discovery Loop, which Google is investing in. The startup has a goal of “automatically solving important problems in machine learning, science, and engineering”.

Hassabis, who shared a Nobel Prize in 2024 for work on using AI for protein folding, has said he thinks we are in what he calls the "foothills of the singularity,” the theoretical moment when Artificial General Intelligence (AGI) sees machine intelligence surpasses human capability, becomes capable of recursively improving its own software and hardware, and triggers an uncontrollable, explosive escalation in technological progress that fundamentally changes human civilisation. In Google's blog post announcing the changes, a typically optimistic Hassabis says, “I’ve been working towards AGI my whole life and now, like many of you, I feel it is close at hand. It’s critical that we collectively get the next steps right to ensure this all goes well for humanity and we usher in an incredible new age of discovery and wonder.”

I honestly struggle to share that optimism, but I did enjoy this video discussion last month from the Royal Society between Demis Hassabis, Alison Noble and Paul Nurse discussing how the scientific community can get the best out of AI.

Meanwhile OpenAI has released a list of ten mathematical problems and computer science questions that it has solved using its latest models. The results were achieved by an internal version of Astra, OpenAI’s next major model. “The total number of tokens needed to find solutions to these problems would cost roughly $2,000 at Sol API rates,” the company says. Although this all sounds very impressive, mathematicians who have looked at the work have come away sceptical, suggesting that the AI has merely "pasted together ideas from the mathematical literature to build a new theorem”.

At a rather smaller scale, a new project from University of Surrey caught my eye. It is aiming to develop AI-powered tools that help countries tackle Schistosomiasis, a widespread tropical disease caused by parasitic worms spread through contaminated freshwater. Schistosomiasis affects millions of people worldwide. Researchers from the Surrey Institute for People-Centred AI (PAI) and School of Veterinary Medicine will combine advanced disease modelling with AI to help policymakers design more effective disease control programmes based on local needs. Led by Dr Joaquin Prada, the project will also contribute evidence to the World Health Organization (WHO) to help shape future guidance on controlling and eliminating the disease. It has been backed by $1.5 million from the Gates Foundation.

Finally, this comic from Brendan Leonard hit close enough to home to make me laugh out loud, and also reminded me why I still write both words and music the old fashioned way.