I published a longer piece this morning on what "AI-first" actually means once you stop treating it as a slogan. The short version. Most businesses do not struggle because of their org chart. They struggle because the same information gets re-explained at every handoff: a call ends, someone updates the CRM, someone else writes the follow-up, and the owner asks again anyway because the record is incomplete. The fix is not another tool, it is one closed loop with five parts — a trigger, rules about what may happen without a human, the tools the AI may touch, a safety check, and a learning step where the process owner reviews and approves improvements. The fifth part is the one that gets quietly dropped, because it is the only one you cannot buy. The part I most wanted to write down is the bit that usually gets left out. Standing up a pipeline is easy. What takes the time is tuning your voice, deciding which channels are worth the effort, and learning the limits of every system you plugged together. So the thing can be built correctly and still sit unused for a while, because you are still reaching for the old process by reflex. That is not a failed implementation, that is what adoption looks like from the inside. Sam Altman said a version of this on David Senra's podcast last month, about the whole economy rather than one business: "The economy just has so much inertia. People keep doing the same things they're doing." Full essay here: https://carinebruyndoncx.substack.com/p/an-ai-first-company-is-not-a-smaller If you read it, the question I would actually like answered: which of the five parts is missing from the loop you are running right now? My guess is it is the safety check or the learning step, because those are the two that feel optional right up until they are not.

Please use this post to provide feedback on the Full v2.5.1 beta release Thank you for helping out.

I've just cut v2.5.1 of ThirdBrain, the Obsidian-based business AI OS I've been building. Two very different asks, and you can take either or both. 🖥️Ask 1: the fresh install v2.5.1 is a release candidate. The gate I can't close alone is the first run on a machine that isn't mine — Windows/WSL, macOS, Linux. Unzip it somewhere clean, open it in Obsidian, point your agent at it, and try to get to your first useful output. That's the whole test. Tell me where you got stuck, confused, or bored. "I didn't know what to click" is a bug report I want. Low commitment — an hour, maybe less. ♾️Ask 2: the self-improving company (looping) - continuously improve on your KPIs (beta) This is the one that needs real testing, and real time. The vault runs a continuous loop over your business — measure → diagnose → act → verify — across your value chains: win and get paid, deliver and keep happy, know your numbers, buy smart, hire and grow. The idea is a company that gets a little better each cycle without you having to remember to make it. Some numbers compute themselves from what's already in the vault (deals, invoices, clients, tasks). Others you type in, because no system can read your bank balance by wishing. Those get flagged stale after three weeks so the dashboard can't quietly lie to you. It's in beta and I don't want to ship it on my own judgment. I need people willing to run it against a real business over several cycles and tell me: - Are these the numbers you'd actually make decisions on, or just the ones that were easy to compute? - Does the diagnose step tell you anything you didn't already know? - Where does it ask for data you don't have? - Does anything actually improve, or does it just describe? Higher commitment — a few weeks of real use. Worth it if you want a say in how it turns out. 💾 What you get: - The full package, not a trimmed demo — one download, one install. Plus every update released during the beta, so you're never testing a stale copy. - The download lives in the skool community and I'll grant you access directly, so I know who's testing what. - Feedback goes in the thread rather than my DMs — you'll learn as much from what other testers hit as from your own run.

Here in Belgium, we are halfway through the summer holidays. This skool community started about 2 months ago, creating the classroom downloads for installation and making more tutorials and reference material available. I am preparing for a new cut of the system this weekend, and plan on doing a kind of office hours weekly meeting for anyone to join, and just hands-on Q&A. So the question is, when would be the best day for this. Weekday or Weekend, and then which time. I think we have people from australia, europe, asia and canada and us (est and pst) timezones. So to start of, if you could reply with your preference(s), we can find the best day and time.

I added a new PDF to the example Website Audit section. It explains the full website audit report pipeline: which skills are involved, which parts come from Starter vs Marketing Free vs QMS Pro, and how the final branded report is assembled. Useful if you want to understand what sits behind the “nice report” output before running your own audit.