creandum.com

The brain stopped waiting to be asked. Agents now run around the clock on a ten minute heartbeat, drafting my replies and repairing the system when it breaks. Nothing they write reaches anyone until I have approved it. A follow-up to A VCs AI Brain, published in June. Same system, same chapters, three months further on. Six building blocks that turn a VC's AI Brain from a search box into a working team: world model, agents, auto-reply, self-repair, apps and security.

Pave Founders, Julie and Jérémy Getting to space is no longer the hard(est) part. Over the past decade, launch costs have dropped dramatically; we are all now fairly used to reading news reports about new rocket launches, or seeing footage of reusable rockets gently landing on launchpads out in the ocean, not to mention the thousands of satellites that are sent into orbit every year. Getting to space is no longer a bottleneck, it’s what you do when you’re there.

The current global AI buildout is the largest and fastest infrastructure project in the history of mankind. But it carries a strange contradiction: the technology powering the most sophisticated software ever built is being bought, sold, and operated like it's 2005. The problem today is that users (organisations) still have to rent whole GPU servers, even when their customers only utilise a fraction of the capacity.

Despite the trillion-dollar scale of the manufacturing industry, many of us still hold an outdated view of manufacturing as low-skill repetition. Modern factories operate under both extreme operational and cognitive load. Production lines move fast, visibility is often limited, workers are spread thin, and even the smallest deviation from plan can trigger hours of downtime, life-threatening safety risks, and tonnes of wasted materials now headed for landfill.

Customer experience is one of the clearest use cases for conversational AI - repeatable processes, narrow scopes, clear resolution distinction -  yet most enterprise deployments are already seeming to stall long before they deliver value. Black-box agents, unpredictable or even untrustworthy behaviour, and rigid legacy workflows have made automation feel risky rather than liberating. Organisations are desperate for efficiency, but they can’t compromise on control and risk customer trust.

One could imagine a world of supercharged legal associates armed with AI tools that accelerate their work. Or perhaps highly interpersonal senior partners commanding agents of legal AIs? Or  Or perhaps companies just buy finished legal artefacts directly from legal AIs? With the rapid developments of LLMs and agents: who really knows. That said, the future does start somewhere. If you ask Henrik & Ludvig at Vesence, that somewhere is in Microsoft Word.

Most organisations live with - or at least have learned to tolerate - a quiet but constant drain on productivity: the manual work that eats hours and hours of employee time every week. Repetitive tasks, brittle legacy scripts, and half-finished automation projects add up to a significant drag on productivity. These are the same manual tasks that many of us believed would be the first things to be solved by the emergence of AI.

“Choose a job you love, and you will never have to work a day in your life” - Confucius Few quotes better capture the impact that our jobs can have on our lives; they determine how we spend our time, what we learn, and who we become. For many of us, our jobs are a path to fulfillment, to satisfy our innate desire to spend our time on meaningful things. The same is true for companies: every hire reshapes what is possible, goals fall in and out of reach with every good and bad hire.

A retrieval system built around the bottleneck of every modern AI: context. The Hook The top users of AI at Creandum cover every position in the firm: GP, EA, Office Management, Legal. The team has been on the AI pill for nearly two years now.I wanted to see how far the current generation of tools can be pushed.

Time catches up with us all - nowhere is this more apparent than in the codebase of an enterprise IT organisation. But there’s a way to turn back the clock: Conduct. Conduct is a London-based enterprise AI company emerging from stealth with one mission: to modernise legacy ERP at scale.