Subhrajyoti Modak

Commented on Supaboard AI · Oct 21, 2026

Glad that's the part that clicked. Yes, that's how the weekly review works: every scheduled run re-queries your live database and writes the summary fresh, so you get what moved and why for that week, not a template with new numbers dropped in. You tell it what the review should cover (for example, what changed versus last week and what drove it), and when you correct a definition, the next Monday's report uses it.

Commented on Supaboard AI · Oct 19, 2026

Hey DevHunt, Subhrajyoti here, co-founder of Supaboard. We spent years as the data people at Fortune 500 teams and startup labs. The job was the same everywhere: write the SQL, build the dashboard, and end the day with one more request waiting. The tools never covered all of it. One answered questions, one drew dashboards, one scheduled reports, and someone still had to glue them together. So we built one agent that does the whole analyst job on your live database: the answers, the dashboards, the Monday report. It shows the SQL behind every answer, learns your definitions from your corrections instead of a semantic layer, and scores itself on questions you already know the answers to. On BI Bench and other accuracy benchmarks, our agents score highest. In a blind test with 134 people, they beat the Claude Code and Codex harnesses. The method is on the site, and I'm happy to walk through it here. What you get: • Query Bench: a SQL and Python editor with tab completion that knows your schema. • No semantic layer to build and no eval harness to set up. Your corrections become rules, and evals run on every plan. • Live on Postgres, MySQL, SQL Server, MongoDB, BigQuery and Snowflake, with no sync. Any tool with an MCP server can be asked questions too. There's a free plan with no credit card, and viewers are always free, so the whole company can read the result. On-Premise with your own keys if data can't leave. We're here all week. Tell us what breaks and what you'd want it to do next. Subhrajyoti