Unlok Insights: AI-written market briefs
Hansraj Saini
AI-written market briefs for retail investors: 16 agents, 9 workflows and a newsroom-style CMS. I lead the project at Kupa, 2026.
The problem
Retail investors drown in market news. Unlok Insights reads it for them and writes short, plain-English briefs: what moved, why, and what happened.
AI writing about money has to be checked. The system needed editors in the loop, versioned prompts and a clear picture of what every model call costs.
What I did
Nine AI workflows and sixteen agents on Mastra produce pre- and post-market briefs, top movers with catalyst detection, and 'What Happened' explainers.
An editorial CMS with a prompt playground, versioned prompt migrations and per-feature AI cost attribution keeps humans in charge.
Whispers on a phone, filterable by date, category, tag and importance
I lead the project and a team of three interns, owning the web app, API, database and the Docker → AWS release pipeline.
Highlights
16 agents in production. Briefs, movers, catalysts and explainers, written by agents, with editors able to edit, regenerate or archive any of them.
Prompts as code. A playground, versioned migrations and model selection, all tracked.
Korean, 2FA and promo codes. Full Korean localisation, sign-in with an authenticator app, and promo codes for launches.
225+ PRs in six months. Shipping weekly while leading three interns.
Where it stands now
Insights publishes AI-written market briefs every trading day at insights.unlok.com, with an editorial CMS behind them.
Stack: Next.js 14, Express 5, Drizzle, PostgreSQL, Mastra, Azure OpenAI, AWS Bedrock. Full case study: hansrajsaini.com/work/unlok-insights
Like this project
Posted Oct 21, 2026
AI-written market briefs from 16 agents and 9 workflows, with a newsroom-style CMS, versioned prompts and AI cost tracking per feature.
Likes
0
Views
2
Timeline
Jan 1, 2026 - Ongoing
Clients
Kupa
