Speech Brief – neutral summaries of public statements
LLM summaries of speeches—but no fact-check, opinion filtering, or integration with analysis tools.

Glassdoor for individuals without the star-rating noise, but moderation will be a nightmare.
Job seekers, hiring managers, and freelancers vetting potential collaborators
Glassdoor · Blind · Klout
This is where I thought we have glassdoor and ambitionbox for companies, but are companies inherently bad or good? No, they aren't. Glassdoor tracks companies, but companies aren't inherently good or bad; they are just legal abstractions. People don't leave companies; they leave bad managers. I have experienced this time and again in my career.
Hence, I started Professional-Health-Ledger as an immutable record of professional conduct. I want the world to at least be net neutral in good vs bad conduct. Hence, you get 1 flag credit for every one vouch you do on PHL. You have to start with a vouch / positive review before you can flag / negatively vote someone.
PHL is non-anonymous; I cannot afford a battery of lawyers to defend fake profiles. To withstand legal scrutiny, the core metric is anchored to a binary, subjective question: 'Would you work with them again?' This is a protected expression of opinion for the public good, not a factual allegation of misconduct. Optional text reviews go through a moderation queue before being committed.
I've structured the platform to respect GDPR/DPDP guidelines, allowing flagged individuals to apply for removal, backed by a transparent audit trail.
Transparency protects everyone. When professional conduct is publicly visible, it creates accountability. Good actors are recognized. Bad patterns become visible.
The stack is simple: JavaScript, CSS, and Vercel, with GitHub Issues acting as the database backend. It's 'vibe coded' and fully open source.
I’m looking for feedback from the HN community on our credit allocation logic, identity-verification graph hardening, and database schema. I'm also looking at how do we verify that an individual 'x' has actually worked with / for an individual 'y' they are vouching or flagging on PHL using some kind of Linkedin based logic.
Live Utility: https://ProHealthLedger.org
GitHub Repository: https://github.com/muglikar/ProHealthLedger
I'll be hanging out in the comments to discuss the architecture, design trade-offs, and edge cases.
LLM summaries of speeches—but no fact-check, opinion filtering, or integration with analysis tools.
Proof-linked STEM work ledger, but solves a problem most builders don't know they have.
Thoughtful personal health dashboard, but it's a blog post—not a tool others can use.
Novel crowdfunding-for-AI concept, but projects show cents raised against hundred-dollar targets.
The product pivots away from pure price-chasing toward risk and market-health signals — think RSI/MACD screeners, an overall fear & greed style health score, private signals and portfolio alerts. The UI looks modern and usable, but the idea lives in a crowded space; to jump from solid to notable it needs transparent backtests, methodology, and clear differentiation on how its "AI-driven" scores actually improve decisions.
Live simulation proves bilateral signed ledgers stop agent scams better than reviews.