Querio

having "cmd + z" in real life would've been sick, don't you agree? not many people talk about it but regretting tattoos as a creative hits you differently. here's my take on it: https://lnkd.in/e2XiKcZ4

The best explanation of your data model ever given happened in a chat session that nobody saved. A practitioner at a large enterprise said this about their Snowflake Cortex rollout: "I think the most important thing are the discussions the subject matter experts are having. I don't know if they're keeping a log of the conversations." Subject matter experts were getting in, doing real work, actually articulating why certain metrics were structured the way they were. None of it was logged. None of it was reusable. When they closed the session, the context closed with it. This is a structural problem. Every insight a domain expert shares with an AI tool is an artifact that should be retrievable. Instead it evaporates. The conversation is the documentation. The documentation disappears. Next week, someone asks the same question again. How many teams have started treating AI conversation logs as a knowledge base? Wondering if this is common practice or if most orgs are still watching it vanish. If you're new to this series, this is my weekly confessions from data teams. I give you one honest confession someone like you has had so you know you're not alone

We've been building agents at Querio for a while now and one thing keeps being true: the more tools, orchestration and boundaries you wrap them in, the worse they get. Remove the layers, give them a machine and some files, and things start to click. I wrote about why software engineering figured this out first, and why data is still catching up. https://lnkd.in/eAXhgBcz

'holy shit that experience was so crispy i want to keep toggling between these tabs just for that animation induced dopamine hit' that's what Mariia and I have become more invested in delivering as Querio matures. the polish, the minor details that tighten your experience. those things you can't quite articulate but deeply feel. we're building towards that, and it feels goooood. our latest blog post covers some of the fundamentals behind crafting clean, polished UX — and how UI consistency has been crucial to landing that feeling. go read it, link in comments

New product update: Embedded Explore 📊 Give your customers AI driven analytics. Delight your users with the ability to query their own data live. Teams are using Querio to let their customers explore things like pricing, logistics, and product usage directly inside their product. So, what is Explore? It brings analysis and storytelling into the same place, so insights don’t get lost between queries, charts, and presentations. We’re also laying the groundwork for a new chat mode that will make Explore far more accessible to non-technical users. 𝗪𝗵𝗮𝘁 𝘄𝗲 𝘀𝗵𝗶𝗽𝗽𝗲𝗱 𝗿𝗲𝗰𝗲𝗻𝘁𝗹𝘆... -𝗣𝗶𝗰𝗸 𝘆𝗼𝘂𝗿 𝗺𝗼𝗱𝗲𝗹 Use Opus and Sonnet 4.6, or the full Gemini suite, with live visibility into model availability. -𝗦𝗵𝗶𝗽 𝘄𝗶𝘁𝗵 𝗤𝘂𝗲𝗿𝗶𝗼 Embed a full Explore via iframe, or build your own UI with our new v2 API and multi-turn conversations. -𝗕𝗼𝗮𝗿𝗱𝘀 𝗹𝗲𝘃𝗲𝗹 𝘂𝗽 View mode keeps things clean for your audience, while new layout tools give builders control over how the story flows. We’re shipping fast while building toward a more conversational Explore experience, where anyone can explore their data without writing SQL. Drop us a comment or message if there are features you'd love to see next. More soon 👀

I like sharing my blog posts with my girlfriend before I ship them. My latest one caused some controversy out of the gate. Her gut reaction was that “empathizing” and “AI” just don’t go together, they feel contradictory. And it’s true, it is a weird feeling, humanizing something that isn’t human. It’s a fairly common thing we humans do tho, so much so that it has a name: anthropomorphism. She read more and pushed away that “yuuuk” feeling to get to the kernel. It’s not about actually empathizing with AI as you would with a fellow human. It’s about keeping in mind the agent's perspective, which leads you to give better context, clearer direction, more thoughtful prompts. That’s the win. As you can already tell, I wrote a bit about this. What it actually means to empathize with your AI agents and why I think it's one of the more underrated skills you can develop right now. p.s. I'll link it in the comments below

𝗣𝗥𝗢𝗗𝗨𝗖𝗧 𝗨𝗣𝗗𝗔𝗧𝗘: 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗧𝗮𝗯 📈 We’ve upgraded 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 (previously Knowledge Base). The goal: clearer structure, better control, and more transparency in how Querio understands your warehouse. 𝗪𝗵𝗮𝘁’𝘀 𝗻𝗲𝘄... -𝗘𝘅𝗽𝗹𝗶𝗰𝗶𝘁 𝗷𝗼𝗶𝗻𝘀 Define joins directly so queries follow your intended data relationships. -𝗖𝗼𝗺𝗽𝗼𝗻𝗲𝗻𝘁𝘀 Metrics and examples now live in reusable components. Cleaner setup. Easier governance. -𝗔𝗱𝗱 𝘁𝗮𝗯𝗹𝗲𝘀 𝗶𝗻 𝗼𝗻𝗲 𝗰𝗹𝗶𝗰𝗸 Browse your schema and add tables to Context instantly. -𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝘂𝘀𝗮𝗴𝗲 𝘃𝗶𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝘆 See how much context each conversation is using. 𝗔𝗹𝘀𝗼 𝘀𝗵𝗶𝗽𝗽𝗲𝗱... 1)MotherDuck + ClickHouse integrations 2)Model selector in Explore (Claude Opus 4.6, Gemini 3 Flash) 3)Granular permissions for editing Context 4)Scheduled refresh keeping your board fresh 🔄 We’re shipping fast to give teams ⏩ speed & control. Next up: Embedded + Slack 👀

I led Texas Guadaloop to Elon Musk’s competition awards, I did my part to help Glovo past Unicorn status, and none of it really comes close to building a startup… In very much Elon style, I slept in the the office for a month straight not too long ago. (the london housing market is to blame tho 😅 but it was a great way to get shit done) With startups, you go all in: - I live with my co-founder - Work 6/7 days a week every week - Do long-distance with my girlfriend It’s intense, but it means that every day, I get to work on a problem I deeply care about. Helping people be data-first, encouraging them to find their passion for data and uncover the superpowers it brings. I didn’t start technical, now I’m CTO at Querio: an agentic intelligence layer that instantly pulls accurate business insights from your database. Everyone wants data-driven teams. Nobody gives them the tools to get it done. And no, prompting ChatGPT and having DBeaver + a postgres user isn’t going to solve it. That’s like having your engineers write production Python in… IDLE. I’m going to make an effort to post more as we scale. Follow along to see how we make every fast-growing data and product team 10x more productive.