Hacker News

Hey HN.

I’m a 17-year-old developer, and I recently needed to compress a lot of images for a small website I was working on.

Most tools I found required uploading files to their servers.

That made me uncomfortable, especially since many of the images belonged to clients.

I also ran into another issue: many tools were slow at compressing multiple images because of server uploads.

So I decided to build something that works differently.

I built ConUtil, an image toolkit that runs entirely in the browser.

Ahmed

Comments URL: https://news.ycombinator.com/item?id=47275090

Points: 2

# Comments: 1

Article URL: https://www.researchgate.net/publication/389855619_Global_Warming_has_Accelerated_Significantly

Comments URL: https://news.ycombinator.com/item?id=47275088

Points: 2

# Comments: 0

Article URL: https://arxiv.org/abs/2512.05887

Comments URL: https://news.ycombinator.com/item?id=47275074

Points: 2

# Comments: 0

Article URL: https://nesbitt.io/2026/03/04/package-managers-need-to-cool-down.html

Comments URL: https://news.ycombinator.com/item?id=47275073

Points: 3

# Comments: 0

Article URL: https://github.com/andrew/gitlocal

Comments URL: https://news.ycombinator.com/item?id=47275057

Points: 2

# Comments: 0

Article URL: https://press.asimov.com/articles/dna-sequencing

Comments URL: https://news.ycombinator.com/item?id=47275049

Points: 3

# Comments: 0

Article URL: https://www.bbc.com/news/articles/cjd98091g28o

Comments URL: https://news.ycombinator.com/item?id=47275035

Points: 32

# Comments: 10

Article URL: https://blog.icme.io/ai-agents-can-move-money-lobstar-wilde-proved-they-can-lose-it-too/

Comments URL: https://news.ycombinator.com/item?id=47275030

Points: 2

# Comments: 1

I’ve been exploring the NPPES dataset, the federal registry that assigns NPI numbers to every healthcare provider in the U.S. It currently has about 9 million records and grows by ~30k per month, but accessing it usually means downloading multi-gigabyte CSVs or using the CMS lookup that returns one provider at a time.

I built NPIScan to make the dataset browsable. You can search by name, NPI, specialty, or location and drill down from state → city → ZIP code. Each provider has a profile with credentials, practice locations, taxonomy codes, and digital health endpoints.

A few interesting patterns from the data:

- 2025 had ~631k new NPI registrations, the largest jump on record

- Behavior Technicians grew to ~526k providers and are now among the largest specialties

- California alone has ~1.1M providers (~12% of the country)

- Only ~0.5% of providers have registered digital health endpoints

Tech stack: Next.js, PostgreSQL, Meilisearch, Redis. The main challenge was making 9M records feel fast to browse. I solved it with denormalized listing tables, Meilisearch full-text search, and Redis caching for aggregated queries. Most pages respond in <40ms after cache warmup.

Curious to hear feedback from anyone working with healthcare data.

Comments URL: https://news.ycombinator.com/item?id=47275022

Points: 1

# Comments: 0

Article URL: https://www.bloomberg.com/news/articles/2026-02-25/hacker-used-anthropic-s-claude-to-steal-sensitive-mexican-data

Comments URL: https://news.ycombinator.com/item?id=47274548

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/watch?v=FzwyGrNpOMU

Comments URL: https://news.ycombinator.com/item?id=47274514

Points: 1

# Comments: 1

Article URL: https://gestrada.dev/posts/bloom-filter/

Comments URL: https://news.ycombinator.com/item?id=47274506

Points: 1

# Comments: 0

Article URL: https://blogs.windows.com/msedgedev/2026/03/05/making-keyboard-navigation-effortless/

Comments URL: https://news.ycombinator.com/item?id=47274502

Points: 1

# Comments: 0

I build this tool for using port forwarding with mosh (if too lazy to open my vscode). Hope it is helpful and hear your help on further improving it.

Comments URL: https://news.ycombinator.com/item?id=47274498

Points: 1

# Comments: 0

Hey HN, I built geo-lint — an open-source linter for content (Markdown/MDX) that checks 92 deterministic rules across SEO, GEO (Generative Engine Optimization), content quality, and technical issues.

GEO is the idea that AI search engines (ChatGPT, Perplexity, Claude) cite content differently than Google ranks it. Things like question-formatted headings, FAQ sections, entity density, E-E-A-T signals, and citation-ready statistics all matter for whether an LLM will pull from your content. geo-lint has 35 rules specifically for this.

The interesting part is the lint loop. It ships as a Claude Code skill — you run /geo-lint audit and it spawns parallel subagents, one per file. Each agent reads the violations, edits the content, re-lints, and repeats until clean (max 5 passes). The linter is fully deterministic (no LLM in the rules themselves), so the agent gets unambiguous violation + suggestion pairs to act on. Zero hallucination risk in the analysis layer.

It also works without Claude Code — npx geo-lint --format=json gives you a flat JSON array any agent (Cursor, Copilot, Windsurf) can consume. The rules are the same either way.

MIT licensed, zero runtime deps beyond gray-matter. npm: @ijonis/geo-lint

GitHub: https://github.com/IJONIS/geo-lint

Comments URL: https://news.ycombinator.com/item?id=47274490

Points: 1

# Comments: 1

Article URL: https://createadaptablelife.com/2026/03/how-can-we-stop-waiting-to-do-what-we-say-matters-to-us.html

Comments URL: https://news.ycombinator.com/item?id=47274484

Points: 1

# Comments: 0

I’m trying to sanity check an idea with people who work with regulated systems.

In many organisations we have scanners for code, monitoring for systems, and multiple layers of controls and audit. But the business process itself is rarely checked for logical vulnerabilities before it goes live.

Processes like KYC onboarding, approvals, payments or compliance workflows are often designed in meetings and documented later. Over time more controls get added and monitoring improves, but the underlying process logic is rarely tested.

Which raises a simple question: can this process be bypassed?

I started experimenting with describing processes as state machines and running static checks on them. Things like reachability, missing review steps, irreversible actions without compensation, and similar structural issues.

The idea is to detect what you might call "business process vulnerabilities by design" before the process is deployed.

The page explains the concept and shows a small prototype. The prototype lets you describe a process as a state machine and run automated checks against rule sets (for example operational risk or resilience rules).

What I’m mainly trying to understand is whether this is actually a real problem in practice.

For people working in fintech, banking, risk or operations:

Do process bypasses or design gaps show up in real systems? How are new processes usually reviewed before they go live? Where do things tend to break down? Paper: veilgovernance.com/research/missing-first-line-of-defence

Comments URL: https://news.ycombinator.com/item?id=47274481

Points: 1

# Comments: 1

Article URL: https://www.pewresearch.org/religion/2026/03/05/in-25-country-survey-americans-especially-likely-to-view-fellow-citizens-as-morally-bad/

Comments URL: https://news.ycombinator.com/item?id=47274464

Points: 2

# Comments: 0

Article URL: https://apnews.com/article/daily-telegraph-axel-springer-paper-buy-german-1386ec56fef373bf5ab5b0f7226bf7bb

Comments URL: https://news.ycombinator.com/item?id=47274453

Points: 2

# Comments: 0

I built DiffDeck because I was struggling to review larger pull requests in GitHub, especially ones with a lot of AI-assisted code.

GitHub's diff view works well for smaller changes, but once a PR gets big I usually want more of an editor-style workflow while reviewing ie see the surrounding code, jump to related symbols and files, and mark off what I have already reviewed and I felt Github's interface was really frustrating me.

DiffDeck opens a GitHub pull request in a review workspace with:

- full file context - go-to-definition and references for TS/JS - review notes - per-file reviewed state and review progress - hide/checkoff reviewed files

One thing I wanted was for it to feel closer to VS Code than a traditional PR tool. You can jump around the codebase while reviewing, and features like go-to-definition are meant to feel familiar if you already spend most of your time in an editor.

Right now it requires GitHub sign-in, because the point is to open pull requests you already have access to and review them with more context than GitHub's diff view gives you. I considered making a public demo, but that felt less representative than letting people try it on their own PRs.

This is an early alpha. Right now the code navigation features are focused on TypeScript and JavaScript codebases. The main thing I'm trying to learn is whether this is actually a better review workflow than staying in GitHub's PR UI. For now you can feel free to review a single PR.

I'd especially like feedback from people who review large PRs or AI-generated code:

- what still feels missing - whether this solves a real problem or just one I personally had

Comments URL: https://news.ycombinator.com/item?id=47274437

Points: 1

# Comments: 0