Blog
2026-11-16Presentation
What if your business teams built the data platform? (Forward Data Conference 2026)
The slides of my talk at the Forward Data Conference: at Optic 2000, we put conversational AI at the start of the chain, in the code editor. The result: 25 contributors, most of them business users, build the data platform.
2026-11-16
Self-Service Analytics: Why Your Business Teams Shouldn't Just Be AI Consumers (But Builders Too)
Everyone wants conversational AI to consume data. At Optic 2000, we also put it at the start of the chain, in the code editor: 25 contributors, most of them business people, build the data platform. The full story, with links to the articles that detail each part.
2026-11-13
The Platform That Learns: Every Review Comment Becomes a Rule
At Optic 2000, four people can't review everything 25 contributors and their agents produce. Review holds because a correction is only made once — it becomes a rule — and because a Pull Request arrives already reviewed by an agent before a human opens it.
2026-11-12
A Dashboard Shows, a Data App Acts: What Our Business Teams Build on the Data Platform
At Optic 2000, we thought Streamlit would be used to display charts. Business teams turned it into something else: tools that replace SaaS products, front-ends on top of SAP and Salesforce, embedded agents. What these data apps do, what they read and what they write, and what keeps them consistent.
2026-11-05
Boringly Simple: How 25 Contributors Change Our Data Platform With Almost No Review
At Optic 2000, 25 contributors and their AI agents change the data platform with almost no review from the platform team. The secret is a boringly simple platform: configuration separated from deployment glue, settings derived rather than declared, security and monitoring built in, and rules that reach the agent where it works.
2026-11-04
One Name, One Definition: How We Document 1,800 Columns Without Documenting a Single dbt Model
At Optic 2000, no dbt model describes its columns, yet every column has a definition: it is attached to the column's name and written once, in a business dictionary. Naming conventions, guardrails for coding agents, a generated data portal: how a single way of naming things eliminates definition drift.
2026-11-03
Nobody Codes by Hand: How We Went from 4 to 25 Contributors on Our Data Platform
At Optic 2000, nobody writes code by hand: in ten months, the number of contributors to the data platform went from 4 to 25, most of them business people. The BI team first, then business teams at headquarters: how we onboard them, what happens when one of them leaves, and what it changes for the data team.
2026-11-02
Independence at Optic 2000: Zero Proprietary BI, Interchangeable Infrastructure and AI of Your Choice
Changing cloud, BI or AI model without rebuilding everything: how we decoupled Optic 2000's data platform from its vendors, keeping all business logic in a single repository.
2026-07-03
The Ticketing 2.0 Trap: Why the 2026 Data Platform Must Not Separate Curation and Analysis
The data team prepares, the business analyzes — it sounds idyllic. In reality, it's a golden trap that reproduces the exact organizational flaws we've spent ten years trying to destroy.
2026-04-17
Why Claude Code in VSCode is a Game Changer for Data Analysis
Claude Code in VSCode is the best tool for deep-dive data analysis — ending shadow data, shifting left, and building the platform as you explore.
2026-04-10
Why I Left Managing 100+ Engineers to Build From Scratch Again
After 6.5 years at Nickel scaling from 3 to 40+, I chose independence, hands-on building, and the AI era.