“Basedash is the first tool that writes and runs SQL, catches its own errors, self-corrects, and genuinely nails it.”

Lex is an AI-powered writing tool used by over 300,000 writers, including teams at Google, Microsoft, and Stanford. Founded by Nathan Baschez — previously VP of Product at Substack and Head of Product at Gimlet Media — Lex helps writers produce better work with AI that acts as a collaborator, not a replacement.

As a four-person team with hundreds of thousands of users, Lex needs to move fast on product and growth decisions. But answering the questions behind those decisions — which features are driving retention, where are costs spiking, what does usage look like across cohorts — used to require hours of manual SQL work.

“It was always high effort. Now, with Basedash, it just feels like chatting with the database directly.”

Before Basedash

Lex’s analytics workflow was typical for a lean startup:

• Blazer, a self-hosted tool for saved queries and basic dashboards
• Hand-written SQL, sometimes exported into Google Sheets for visualization
• Claude for help generating queries — but still requiring heavy prompting and manual debugging

For a founder who needs to investigate product questions daily — user engagement patterns, cost breakdowns by AI model, trial conversion funnels — the overhead was constant. Every question meant context-switching away from product work into query writing and debugging.

Why Basedash

Nathan had been looking for something that worked more like a colleague than a tool:

• Write and debug SQL automatically
• Handle errors without human intervention
• Deliver results that could be explored further in real time

That is what he found in Basedash. With Lex’s Postgres database connected, Nathan had a conversational interface over the same production data powering the writing platform — no dashboard setup or schema documentation required.

“It feels like collaborating with a colleague, not a BI tool. It’s performing at the edge of what’s possible with LLMs right now.”

Instead of preparing dashboards or wrangling queries, Nathan now runs iterative conversations whenever he needs to investigate a problem or validate an idea — the kind of exploratory, follow-up-heavy analysis that traditional BI tools make painful.

“The natural paradigm for data is chat. Basedash makes that feel real.”

A critical incident on the road

While traveling in San Francisco, Nathan received an alert: Lex’s AI costs were spiking well beyond normal levels.

In the past, resolving the issue would have required writing SQL queries on a laptop, or worse, on his phone.

“I’ve written SQL on my phone before in Blazer. It’s not a pleasant experience. Basedash made it simple.”

Instead of fumbling with scripts, Nathan opened Basedash and asked:

“AI costs are spiking — can you figure out what’s going on?”

Within minutes, Basedash guided him through the same thought process he would have taken himself: was the spike coming from a small group of users, a particular model, or a specific feature? The tool asked clarifying questions along the way.

The diagnosis: bot users were abusing Lex’s free trial to run model distillation attempts. Because he could identify the root cause quickly, Nathan and his team patched the vulnerability before costs escalated further.

“Basedash turned a very stressful moment into a simple fix.”

Impact

For a small team building a product used by hundreds of thousands of writers, every hour counts. Since adopting Basedash, Nathan has replaced a patchwork of query tools with a single conversational interface:

• Investigative questions that took hours of SQL and debugging now take minutes
• Cost and usage anomalies get surfaced and diagnosed in real time, from any device
• Product and growth decisions are backed by traceable, shareable conversations instead of one-off queries
• Zero time spent building or maintaining dashboards or internal tools

“Basedash feels like collaborating with a colleague, not a tool. It’s the future of how teams will work with data.”