Co-founder & CEO @lightdash_devs (YC S20), building the open-source agentic data platform to help teams make sense of their data.
London
Joined April 2011
- Tweets846
- Following1K
- Followers583
- Likes2K
Pinned Tweet
90% less tokens, 80% faster. That's it - that's the message.
Six chart follow-ups took 5.5s with Fast mode and 59.8s without.
Watch Lightdash’s Jev-powered fast path in this side-by-side test. @joaovianacodes shares how he built it (and what broke) in our newest blog: go.lightdash.com/e986e9
Why do we still have per-seat pricing in 2026??
A lot of data leaders I talk to have had to ration licenses at some point. But in an agentic world where people aren't even using UIs, it's an archaic model.
SaaS companies need to move towards flat pricing. It's one of the reasons teams often give us for switching. The whole job of a data team is getting more people to look at the numbers, and your pricing shouldn't punish you when it works.
If you're still paying per seat, come chat to us!
Congrats @JohnHealey_MP - stifle growth and push away entrepreneurship at any cost in the UK 👏
Have you seen the UK government's new Founder Relief proposal? No CGT on your first £10M on an exit. Sweet! But read it in detail. Your company is unlikely to qualify, and you should assume that it will not apply to you.
To recap: Once upon a time, there was Entrepreneurs Relief. If you had more than 5% equity in a company, then you paid only 10% capital gains tax when you sold your shares, up to a lifetime value of £10M. That was sweet!
Then the government changed the rules, renamed it to BADR (though Worser would have been more appropriate) and cut the lifetime limit to £1M.
In today's budget, they've reduced the BADR limit to just £500k - which as a tax benefit is minuscule.
To compensate, they brought out a fabulous-sounding new Founders Relief - consultation paper link below.
When you start reading it, it looks great. It applies to founders who've got more than 25% equity at any point in the company's life, which probably means if your company's got 3 founders or fewer, you probably qualify.
But then it goes into such a convoluted maze of growth, revenue, and R&D requirements that few companies are ultimately going to qualify.
And that's really the thing that makes it useless, because the whole point of such a relief is that, as a founder, when you're weighing up the pros and cons of building a business in the UK versus, for example, moving to the US, you see at the outset that, great, I'm going to pay only 10% CGT on my shares. But here, there is so much uncertainty that there's no way you can bank that as a benefit, and so ultimately it fails its key purpose, which is to drive decision-making for an early-stage founder to build and stay in the UK.
So, fabulous work by Startup Coalition and Dom Hallas on getting this and the other package of founder benefits that they've managed to get into the budget at a time when that must have been an incredibly hard uphill climb with the government.
I'd love it if this benefit was something that founders could bank on at the outset, but it's not. So I'm looking forward to using the consultation period to turn this into something deterministic rather than having the same chance as a lottery ticket.
assets.publishing.service.go…
One of our founding product principles at @lightdash_devs has always been that everything must be as-code.
I come from a software engineering background, and @datumnite and I wanted to build a tool that felt as good as a software engineer's toolkit. So Lightdash had to work that way; every other analytics platform relied on UIs and left people stuck in the clickops world.
Agentic coding has been the ultimate tailwind for us - in the past 12 months, >60% of building in Lightdash has been done via agents in Cursor, Claude Code, Codex, etc. This is only possible because of the decision to make code assets the "primitive" of the product. Other competitors have slapped on MCPs or opened up APIs, but that doesn't replace building every single pixel as a code primitive from the ground up.
Our decision wasn't the obvious one to take back in 2021, and now it's fast becoming the default for new and legacy vendors who are rebuilding their stacks - personally, I'm glad to see it, and I think this will accelerate the development of data products 10x over the next year.
Gated content is a gift to your competitors.
For the last two weeks, >20% of people who signed up for @lightdash_devs told us they found us through AI, making it our second-biggest source of inbound.
This has happened because, since day one, we've built Lightdash in the open. The code is open source, the docs are public, and none of our marketing content is gated, even our demo doesn't collect your email - which means it's easy for LLMs to read everything we're doing.
Building in public is a key value for me. Now it's how a lot of people find us.
Don't join a company if the founders aren't still in control; if you see a professional CEO, run for the hills.
We added more new customers and revenue at @lightdash_devs in Q3 of 2026 than we did in all of 2025.
We continue to do this the "Lightdash Way", with small, high-performing teams with high levels of autonomy - everybody ships, everybody sells - no exceptions. It's the fastest way to build both your product brain and user empathy.
You can’t be fast and good is fake news spread by the slow
When we first shipped Autopilot, our customers were very honest with us about it. They said it did the job, but it was too trigger-happy. Data teams wanted to decide what it could change, see exactly what it had done and why, and run it on the models of their choice. So we rebuilt it between the public beta and going GA.
Now you choose what Autopilot is allowed to touch, and it can run in read-only mode while you get comfortable with it. Every action is logged, and anything it removes can be restored from the trash.
That's one of my favorite things about building at Lightdash. Our customers tell us what they need, and the team crafts it immediately.
Y’all cooked with this one 👏🏽
Introducing FLUX 3 Max, the latest post-trained video model from fal Research in partnership with @bfl_ai.
FLUX 3 Max is tied for #1 in image-to-video quality on independent third-party benchmarks, generating a 10-second, 720/768p video in ~12 seconds.
For the next three weeks, FLUX 3 Max is 75% off, making it the most affordable frontier video model at its quality level.
We recently launched Autopilot: an agent that maintains your data projects for you. @joaovianacodes put together a short demo showing how it works. Try it out in @lightdash_devs and let me know what you think!
Hamzah retweeted
I spent last week building with Claude Dashboards and they're good.
Claude can use any MCP server as the data source, and if that MCP server can write content, you can push everything you build back. I wrote up the full walkthrough
👉 Read it here: lightdash.com/blogpost/how-t…
Claude launched its newest artifact last week - dashboards! We've been testing them out at @lightdash_devs and they're pretty great.
@datumnite wrote up a short blog on how to use them: lightdash.com/blogpost/how-t…
The future is getting closer. Excited to share this one, the bangers don’t stop. 🚀
