John Ling, co-founder and CEO of Meridian, is building AI tools for spreadsheets, aiming to make Excel workers dramatically faster. He previously worked at Scale AI, where he was regarded as a top 1% performer, and has raised over $15 million in funding, with the seed round led by Andreessen Horowitz. The episode explores his path to founding Meridian, his philosophy on learning and execution, and why he believes spreadsheets are the next major frontier for AI-driven knowledge work.
How John Ling became a top performer at Scale AI
John joined Scale AI because he saw it as a unique vantage point to observe how large language models were developing in real time, and he wanted to deeply understand the trajectory of AI technology before starting his own company.
At Scale, he was not confined to a narrow role; he actively expanded into new areas of the business, driven by curiosity and a desire to learn across domains.
He spent significant time reading research papers and studying data quality across many domains, asking foundational questions like what makes data high quality and what researchers actually care about.
His core responsibility was ensuring the data Scale produced was valuable, which led him to work extensively on benchmarks, evaluations, and internal process improvements using LLMs.
Kimberly Tan of Andreessen Horowitz, who met John through mutual contacts at Scale, noted that he consistently took a first principles approach, was unafraid to voice dissenting opinions, and moved to make changes happen regardless of organizational constraints.
The insight behind Meridian
John noticed a stark contrast between how AI was transforming coding and how it was barely touching finance and spreadsheet work.
In coding, tools like Cursor had compressed tasks that once took weeks into hours or minutes, and the people building these tools were the same people using them, so they had a clear sense of what success looked like and where models failed.
In finance and spreadsheet work, no one had spent the equivalent of 1,000 hours trying to build complex financial models with AI. Bankers still did everything by hand, and even when models produced wrong numbers, non-experts struggled to diagnose exactly why.
He believed the problem was not that AI couldn’t handle spreadsheet work, but that no one had done the deep decomposition of the workflow to understand which parts models could handle well.
This gap represented a massive opportunity: spreadsheets are arguably the most widely used programming language in the world, and the category of Excel-based knowledge work is enormous.
Why spreadsheets are the next frontier for AI
John argued that as AI reshapes knowledge work, there is almost no larger category than the spreadsheet worker, drawing on his own experience as a banker and consultant.
Meridian’s vision is to augment spreadsheet work the way coding tools like Cursor and GitHub Copilot have augmented software development, infusing spreadsheets with meaningful intelligence and automation.
The core thesis is that sustained, deep engagement with AI in a specific domain builds an intuition for what is possible today and what will be possible in months or years, which is an extremely valuable unfair advantage.
John’s philosophy on learning, action, and building
Bias toward action: John’s core belief is that you learn by trying things you have never tried before, and failure is acceptable as long as you gain knowledge from the attempt. He encourages reaching out to anyone, regardless of how impossible it seems, because you genuinely don’t know until you try.
Building an experimentation culture: At Meridian, he fosters an environment where people are encouraged to solve problems their own way, and if they fail, the team supports them rather than punishing them.
Spending 10,000 hours with AI: John advocates spending as much time as possible working with AI in your domain of interest, because sustained exposure builds intuition about current and near-future capabilities that others simply won’t have.
Prompting as a thinking tool: He believes the act of explaining a task clearly to a large language model is itself valuable, because it forces you to clarify what you actually want to do, giving you insight into your own thinking and goals.
