Research boards
What I'm actively exploring
Running notes on the threads I'm pulling on right now. Living documents, not conclusions.
AI-assisted dev workflows
building AI / tooling · Updated May 22, 2026
Pushing on how far agentic coding tools can go when kept on rails — and where self-hosted open models replace paid APIs for the boring, high-volume work.
Running notes:
- Claude Code / Codex / Harness as the front-of-funnel for planning + diffs.
- Self-hosted OSS LLMs for summarization and classification volume — cost to near zero.
- Open question: where does the human-review checkpoint give the best leverage-to-effort ratio?
Identity / security · Updated May 18, 2026
Passwordless and passkey-first customer identity — and the governance layer (Sentinel) that lets security teams enforce policy without adding user friction.
Running notes:
- Passkeys are finally good enough to be the default, not the fallback.
- Governance is the unsexy differentiator: policy, audit, lifecycle.
- Drawing on identity work from the bank / enterprise years.
Short-form news aggregator
building Media-tech · Updated May 16, 2026
An Inshorts-style feed for the Indian market — fast, calm, mobile-first reading backed by an AI summarization and ranking pipeline (shipping as Shawt).
Running notes:
- Bet: people want news faster and calmer than the infinite scroll allows.
- Summarization pipeline runs mostly on self-hosted models.
- Editorial guardrails matter more than raw model quality.
Spiritual-wellness · Updated May 10, 2026
A spiritual-wellness platform combining Vastu, Feng Shui, and astrology, with a prompt-intelligence module that keeps guidance personal and explainable.
Running notes:
- The hard part isn’t the models — it’s encoding traditional frameworks faithfully.
- Prompt-intelligence layer: structured retrieval over curated sources, not freeform generation.
- Validating demand with a small cohort before building more.