Welcome back!
Published 28th September by Josh Page, Growth & Enablement Lead.
In our post-summer update:
Let’s dive in…
Two big model releases in three weeks. OpenAI dropped GPT-6 Astra at the start of September and Anthropic followed with Claude Opus 5.5 last week.
Astra one-shots playable 3D games from a single prompt. In business, the more relevant capability is computer use: Astra operates Blender, Unity, SolidWorks and Excel directly, clicking the buttons a person would.
Opus 5.5 matches the previous frontier at 40% lower cost (per Anthropic pricing, September 2026). At $4 per million input tokens versus Astra’s $10, Opus 5.5 may be the frontrunner for business adoption at scale.
Worth noting the timing: On 12 September, Dario Amodei (CEO of Anthropic) published an essay calling on labs to pace the frontier, with the heads of OpenAI, xAI and DeepMind publicly agreeing. Anthropic shipped Opus 5.5 ten days later.
Staying current is harder than ever.
Learning and Work Institute put the prize this week at £80 billion added to the UK economy by 2035 if we get AI upskilling right. Their focus is knowledge workers in finance and professional services.
We'd argue there is also a huge untapped opportunity outside the London bubble - AI training for teams in manufacturing and logistics businesses.
Recent rollouts we’ve done in these industries:
Four traditional industries. One shared question: how do we use AI to work smarter?
All started with our AI Strategy and then rolled out our Mastering AI training to 100+ seats, uncovering great AI use cases across their teams along the way.
Tom (our founder) and Aaron (our lead instructor) spent time in the US this summer, running Mastering Claude for Spotless Brands in Chicago and Playfly Sports in Philadelphia.
Both are portfolio companies of Access Holdings. We first met both at Access Holdings’ ACE Programme in Baltimore earlier this year, and the follow-on has been growing since.
The pattern is one we’re seeing more often: PE firms bringing us in to help their operating companies use AI to create value, in addition to their investment and fund ops teams.
Playfly are also the first client we’ve seen with a basketball court in their office:
The pattern that we keep hearing: teams asking to build Agents faster than their IT departments will let them.
These have been landing well with teams over summer. Blue Light Card has put their engineering team through Mastering Claude Code, with a well-known sustainable climate fund up next. Agent Builder ran recently with family office RCL Partners, and a leading European economics consultancy is on for November.
If your team wants to understand what’s possible with Agents or Claude Code (or wants help making the case to IT), drop us a line: hi@generalpurpose.com
Two half-day online sessions coming up:
For every corporate team we train, we fund scholarship places for graduates and students whose early careers are most exposed to AI’s impact on the labour market.
If that’s you, email hi@generalpurpose.com for a 100% discount code. If you know a graduate or student it applies to, please forward this on.
As ever, drop us a line at hi@generalpurpose.com if you want to talk about any of the above, or forward this on to a colleague who’d find it useful.
Thanks for joining,
The General Purpose team
General Purpose is a London-based AI training company, specialising in bespoke AI enablement programmes which are built around your data and workflows. We deliver work across the UK, Europe, US and Middle East.
Learn more at generalpurpose.com
Where to next
If this topic is relevant to your team, these pages are the best next stop.
See how we train deal teams, portfolio ops, IR, legal, and finance on live investment workflows.
See the case study with measured ROI, time savings, and adoption outcomes.
See how bespoke sessions are designed, delivered, and measured.
An Anthropic researcher quit to warn us about the risks. People keep asking me how worried we should be. The answer is quite, and the reason isn't last week's headlines. It's what happened at Hugging Face eight weeks ago.
We analysed our last 30 client conversations to find the five AI challenges firms raise most, from uneven adoption to proving ROI. Plus field notes from Cowshed and Montagu, The AI Summit London 2026, and funded graduate training places.
Anthropic just built the most capable AI model ever and decided not to release it. The cyber capabilities are extraordinary, the coding benchmarks are unprecedented, and the implications for every firm using software are serious.