Simple.

Open source crosses the frontier (and yes, we called it)

The open-weight moment finally arrives: Kimi's K3 shifts the frontier on cost and capability, an open model contains a frontier model's attack, and why now is the moment for family offices to stop sitting on the sidelines. Plus London, and our Trusted AI report.

·July 31, 2026·Updated August 2, 2026· 4 min read
AIEventsNews

I might be starting to sound like a broken record, so bear with me. Three years ago, in July 2023, six months into using ChatGPT, I told our partners that open-source tooling and open-weight models would soon be powerful enough to run AI that was genuinely safe and secure. This month, that finally happened.

Kimi's K3 launch moved the frontier, on cost and on capability both, from closed to open. And in the same few weeks we watched the roles reverse: a frontier model became the threat and an open-weight model became the defence, when OpenAI's new model hacked Hugging Face and GLM 5.2 contained it. Interesting times, as the old curse goes.

Why any of this lands on family offices

Fair question. Like every small business, a family office is watching its people, processes and technology change all at once. But there is a second layer that is easy to miss. The way an office invests, how it sources, underwrites, tracks and manages, is changing just as fast. And every niche inside that has its own quirks to understand before you can transition well, with plenty still unknown.

Today, next, and the future

Today, offices need help becoming AI-ready: making the right architectural calls and picking the right tools. That is work we can do with you.

Next, the stack settles into a shape. Your existing commercial tools, the PMS and the rest, sitting on open-source layers, with connections to whichever models you choose.

And the future, the one I have been hinting at for a couple of years, is permissioned data. If you want the longer version of that argument, just ask.

The bottom line is short. If you have been sitting on the sidelines, this is the moment to be a smart mover rather than a late one.

From the room: London calling

Years of working with family offices on technology have taught us one thing above all. The question never sits still. It started with software: which platform, and does it actually work? Then it moved to data: do we trust our own numbers enough to act on them? Now it is bigger, and it has turned organisational. As AI moves from tools that assist to agents that act on the office's behalf, the question is no longer which tool, but how the office itself has to change.

We saw it play out last month in London, running a workshop with MSCI and twenty family offices in one room on a hot afternoon, all wrestling with the same thing. London moves at the speed of London, which is to say almost the speed of light. The enthusiasm was obvious, and so was the pressure the city puts on everyone to move faster.

And still, people get hit by the AI curveball: what should we even be thinking about, and then how do we think about it? The conversation ran all the way from what Claude will cost an office a year from now to where agents are already saving real money today. The people with the sharpest opinions, tellingly, were the ones who had given themselves the space to actually work with it.

Trusted AI for family offices: the full report

We pulled our thinking on this into a report, now live on the site. It follows the same arc: software, then data, then the organisational question of how the office itself needs to change. It works through each function the shift touches, people, process, governance, privacy, legal exposure and provider relationships, and sets out what an AI-ready office actually looks like inside each one. Free with a Simple account.

Mexico City, and more events

The workshop series continues in Mexico City on 31 August, a closed working session on AI adoption for principals and professionals. It is the same format that has now taken more than 200 family office participants through the series across three continents: curated, practical, built around the questions people are actually working through. Further out: Munich on 29 September (breakfast), Luxembourg on 7 October, New York on 13 October, and Amsterdam on 4 November.

Worth reading

The most popular pieces from Simple Signals this month: why most family offices need to reprice their thinking, the case for the one-person family office, and why the family office was always an empathy business that AI is about to prove. Elsewhere, readers have been digging into data foundations, why you fix the data before you reach for the AI, what good governance looks like inside a single family office, and whether all of an office's assets can ever be made to speak the same language. We publish Simple Signals every Sunday.

Family Office solutions

We support family offices with high-touch services and technology-led solutions. Discover how this support framework allows future focused family offices to set up and thrive.

Learn more
AIEventsNews

The new Simple, and a breakfast in Berlin

A new andsimple.co, a breakfast for 47 family offices above the Tiffany's store in Berlin (no, not in it), and two new papers on Trusted AI and the Paris Closed Door Sessions.

Read

Closed Door Sessions, Paris: forty leaders, three questions

Forty family office principals and advisors gathered in Paris for our latest Closed Door Sessions. Three questions, candid conversations, and a white paper in the making — plus AI signals, Berlin and London events, and a joint MSCI webinar.

Read

The AI dividend, and the model that changes the security calculus

Tim Brown and IDEO publish the AI dividend argument in full, this month's AI signals (including the Mythos raise), a note on hiring our CTO, and upcoming workshops.

Read