A personal agent made me want a better Claude.
I tried Instinct, one of the best known personal agents. It was very good, but there was nothing it could do that a well set up Claude couldn't.
Hi folks,
This week I'm stepping back from the news a little and sharing what I found using a personal agent myself.
What we're seeing.
I tried a personal agent. It made me want a better Claude.
Two weeks ago we mentioned that chief of staff agents are all the craze. These are personal agents: you text or call them with a task and they go off and do it. Instinct is one of the best known. Last month it raised US$1bn at a US$10bn valuation, from Sequoia, Benchmark and Coatue, still invite-only (Reuters). So this week I tried it.
What I used it for:
- Organising my emails daily. It did this well, but Claude can do it too.
- Saving restaurants to my Google Maps while I was travelling. Nice to have, not something I would pay for.
It was very good. But it made me realise I just need to improve my Claude. There wasn't anything it could do that Claude couldn't, if I set Claude up well. For me the biggest value from AI is coding related, and that is something Instinct can't do. I also didn't love texting it. It felt a bit too human-like for me, and it is not that hard to open Claude when I need something. I didn't need another thing to be testing, and I couldn't think of anything it would do beyond what I already had.
So I'm not convinced by this market yet. I think a lot of people are finding these tools make it easier to do what Claude and ChatGPT can already do. That is real value, but it is a convenience layer, and the labs are building the same convenience into their own apps every few weeks.
What this means for a fund:
- Set up what you already pay for before buying something new. Most teams have Claude, ChatGPT or Copilot licences sitting at a fraction of what they can do. Connecting them to your email, calendar and documents, and writing down how you like things done, gets you most of the way.
- Every new agent is another thing with access to your inbox. Each one needs its own security review, its own permissions and its own answer to "what did it do?" Fewer tools means fewer of those conversations.
- Ask the vendor what it does that the lab can't. If the answer is "it is easier to set up", that gap tends to close fast.
In the mix.
- Meta is back in the frontier race (TechCrunch)
- On Monday Meta released Muse Wonder, the first model it calls frontier class. Meta says it matches OpenAI's GPT-6 Astra and Anthropic's Claude Fable 5.1 on benchmarks.
- On Thursday it followed with Muse Spark Open, an open-weight model you can download and run on your own hardware, under Meta's own licence (VentureBeat).
- Our read: I'm interested to see how Meta re-enters the game and whether it can be a real contender. A model on par is one thing; getting into the tools people already use every day is the harder part.
- US bank regulators ask: can you switch your agents off? (Reuters)
- Reuters reports the Fed, OCC and FDIC are asking banks what limits sit on their AI agents, whether they can track and audit an agent's decisions, whether humans stay in key decisions, and how they would shut an autonomous system down.
- Why it matters: these are the same questions we hear from funds every week. If you are putting agents near fund data, have the answers written down before your auditor or LPs ask.
Readings.
If you want to double click into things this week I'd recommend:
- AI at work: Evolution from System of Record to System of Work, Tidal Ventures (Tidal). For decades enterprise software was built around the database: the CRM, the ERP, the place the data lives. Tidal argues the next era belongs to software that does the work itself, embedded in the workflow, picking up the know-how people use to get tasks done. It is a good frame for what I found with Instinct: the value is not in another app, it is in the work.
Levercon builds custom, self-improving AI agents and systems for investment firms, with every output traced, verified and signed off.
Book a strategy call.
See where AI fits in your firm.
A short conversation about where AI could help your firm, what to build first, and what it would take.