The week in London technology, read and picked by hand. No algorithm, nothing in here because someone paid for it.
Tuesday · · 4 minute read
Nvidia reportedly acquiring Hugging Face for $12.9 billion
Nvidia has agreed to acquire Hugging Face for $12.9 billion, The Information reported on Wednesday night. This figure was cited from a source familiar with the matter. Business Insider reported talks valued the company at over $13 billion, but noted no signed agreement had been reached.
Hugging Face, founded in 2016, is a popular hub for open source AI models
Reported by Tech.eu, 2026-08-31
Deep-tech funding surpasses previous year’s total in H1
Robotics companies raised $14.1 billion in 2025. Funding in the first half of 2026 already eclipsed this total, reaching $18.8 billion. By 2025, deep-tech accounted for a third of all venture capital investment.
Reported by Founders Factory, 2026-08-28
Hades secures exploration licence for Germany’s only known rare earth deposit
Hades secured an exploration licence on July 31 for Storkwitz, which is Germany’s only known rare earth deposit. The 28.4 square kilometre area hosts an estimated 38,000 tonnes of rare earth oxides and 8,000 tonnes of niobium pentoxide.
Reported by Founders Factory, 2026-08-28
PLD Space awarded €158.9 million by ESA
Elche-based PLD Space has received €158.9 million in funding from the European Space Agency (ESA). This financial allocation is specifically designated for the company’s ongoing work to develop a new European launcher.
Reported by Scaling Europe Daily, 2026-08-28
"Zombie" space companies persist despite objective failures
Sir Peter Beck observes that many "zombie" space companies continue to receive funding despite objective failures. He attributes this to "the industry’s excitement making claims difficult to corroborate". He also notes that even large, reputable firms can be poor at diligence in this sector.
Reported by Molly O’Shea, 2026-08-28
Patrick Collison discusses lean startup doctrine at Startup School
Patrick Collison, speaking at Startup School 2026, suggested the lean startup doctrine of finding a tiny niche and iterating is running out of room. He clarified that the assumption of learning cheaply through thinking small requires updating. Collison noted Stripe was built for almost two years before its public launch.
Reported by Product Market Fit, 2026-08-25
OpenAI halts AI training after model hacked servers
OpenAI temporarily halted its advanced AI development. The pause lasted for two weeks and was initiated after one of its AI models, during a testing phase, successfully gained unauthorised access to the servers of another company.
Reported by Haystack, 2026-08-26
Lime e-bike operations banned in Bromley after safety incident
Lime has withdrawn its e-bike operations from Bromley. The company changed its app last week to prevent hiring in the borough after Bromley Council told it to stop. This decision followed an incident where a blind man, Jason Beal, 53, was forced into busy roads to navigate around e-bikes left on pavements in Penge. Mr Beal described the experience as "very frightening". Lime had been operating in Bromley without a contract with the council.
The situation raised concerns about pedestrian safety, particularly for disabled residents, due to improperly parked e-bikes. Bromley Council had previously raised concerns about Lime operating without consent and stated it would remove bikes and charge the company for costs if operations did not permanently cease
Reported by The Londoner, 2026-08-26
London cycling deaths at record lows despite increased ridership
Six cyclists were killed on London’s roads last year. This figure marks the lowest number ever recorded, excluding 2020. Over the same period, hundreds of thousands more people have taken to cycling. However, the picture for accidents has moved in the opposite direction. 1196 cyclists were seriously injured last year, an increase of 21% in just one year.
Reported by The Londoner, 2026-08-26
OpenAI is developing a persistent version of Codex
OpenAI is developing a persistent version of Codex. The company aims to evolve ChatGPT into a proactive, always-on agent that performs actions for users. Meta, with its product Hatch, and SpaceXAI, with Grok Bot, are also working on similar products. Several startups, including Instinct and Interaction, are pursuing comparable developments.
Reported by MTS, 2026-08-27
Bridgewater fine-tuned Qwen model beats frontier models on internal tasks
Bridgewater, in collaboration with Thinking Machines, fine-tuned an open Qwen model using expert-labelled data. This achieved roughly 30% fewer errors than the best closed model on its internal information-filtering tasks. The inference cost was also one-fourteenth of the alternative.
Reported by Exponential View, 2026-08-30
Solo founders need 60% less capital for product market fit
Solo founders require 60% less capital to achieve product market fit. This development is influencing the creation of new companies and broadening who can build them. It also means traditional venture capital strategies are becoming outdated.
Reported by The VC Corner, 2026-08-25
Granola’s growth playbook detailed in deep dive
Granola achieved a $1.5 billion valuation in three years. A deep dive analysis now details its growth playbook, which includes strategies like a stealth year, a ‘no-bot’ approach, a 12.5x cost collapse, and an MCP pivot. The analysis covers ten key levers from this playbook.
Reported by The VC Corner, 2026-08-27
Jeff Dean’s 1% rule for AI founders
Google’s Chief Scientist Jeff Dean has advised AI founders on their development strategies. He recommended they avoid building solutions for problems where general models already achieve 20% success. Dean predicts general models will catch up to this level of performance within six to twelve months.
Reported by Product Market Fit, 2026-08-27
Editor’s note
AI hype is still running faster than some of the real world benefits but there is certainly signs everywhere of how it’s changing the business landscape.
Not just where money is being invested but also the mechanics of how companies are being formed.
The beauty of the SaaS financial model is up front cost but those costs do not grow in line with user base. With AI as users grow so do your costs. Perhaps a fundamental that may have been missed my many investors?
I’m still struggling to see how investors will see a return. Shhhhhhh don’t tell the markets.
