Hacker News

https://x.com/adammajcher20/status/2024221836053995944

Comments URL: https://news.ycombinator.com/item?id=47066060

Points: 2

# Comments: 0

I use Claude everyday (no joke) and kept getting annoyed by the same thing: selecting text from responses with the mouse. Overshoot, re-select, copy, click input, paste. Especially bad in long conversations where you want to reference something from 30 turns ago. asdPrompt is a Chrome extension that adds hint-based navigation (like Vimium) to AI chat interfaces. Cmd+Shift+S activates the overlay, hint labels appear next to every text block. Type a letter to select a block, then keep typing to drill down: block → sentence → word. Enter copies, or you can press an action key (e, d, x) to inject a follow-up prompt ("elaborate on [selection]") directly into the chat input.

Works on claude.ai, chatgpt.com, and gemini.google.com. Adapts to light/dark themes. Free. Built the initial MVP in 2 days using Claude Code — the adapter architecture, NLP segmentation pipeline, and Playwright test harness would have taken a month without it.

Tech details for the curious: site-specific DOM parsers behind an adapter interface, text segmentation via compromise.js with regex fallbacks for technical content (paths, camelCase break NLP libraries), bounding rectangles calculated via Range API + TreeWalker, overlay isolated in Shadow DOM. Tested with Playwright visual regression.

The landing page has an interactive tutorial where you can try the full drill-down mechanic without installing. Happy to talk about the implementation.

Comments URL: https://news.ycombinator.com/item?id=47066047

Points: 2

# Comments: 0

Article URL: https://www.omacon.org/

Comments URL: https://news.ycombinator.com/item?id=47066043

Points: 2

# Comments: 0

Article URL: https://www.sciencedirect.com/science/article/pii/S1385894726005565

Comments URL: https://news.ycombinator.com/item?id=47066041

Points: 2

# Comments: 0

Article URL: https://nocomplexity.com/documents/reports/SimplifyIT.pdf

Comments URL: https://news.ycombinator.com/item?id=47066038

Points: 2

# Comments: 1

i have noticed that Opera tends to get a surprisingly negative reaction in technical communities. but i have not yet seen a clear and technically grounded explanation that justifies the level of dislike. from what i can tell, common criticisms seems to be that the company was acquired by a Chinese consortium or a questions about privacy.

what are your strongest technically accurate arguments against using Opera in 2026?

Comments URL: https://news.ycombinator.com/item?id=47066021

Points: 2

# Comments: 2

Hi HN,

I built Sniptail — an open-source bot that lets you run AI coding agents (Codex / Copilot CLI) against real repos directly from Slack or Discord.

Instead of AI living only inside one developer’s IDE, Sniptail turns it into a team-facing layer:

- Mention the bot and ask how a feature works - Run commands to ask, plan or implement collaboratively - Generate repo reports - Open lightweight PRs

All against explicitly approved repos.

Flow:

Slack/Discord → job queued in Redis → worker clones repo → agent runs → report or PR link posted back in-thread.

It’s self-hosted, repo-scoped, and modular (chat channel, agent, git provider are separate layers).

Not trying to replace Copilot/Codex/CC, the worker runs it on your behalf.

Repo:

https://github.com/Justkog/sniptail

If the project gains traction, the idea is to support:

- more agents (Claude Code, Open Code, …) - more channels (Telegram, Microsoft Teams, …) - more git services (Bitbucket, Azure Repos, …)

Feel free to try it, let me know if you find it actually useful!

Happy to go deep on architecture if helpful.

Comments URL: https://news.ycombinator.com/item?id=47066016

Points: 2

# Comments: 0

Article URL: https://www.ponylang.io/blog/2026/02/teaching-claude-to-write-pony/

Comments URL: https://news.ycombinator.com/item?id=47066011

Points: 2

# Comments: 0

Article URL: https://www.dailymail.co.uk/health/article-15571317/Urgent-recall-trendy-supplement-moringa-salmonella-deadly-strain.html

Comments URL: https://news.ycombinator.com/item?id=47066008

Points: 2

# Comments: 1

Hey HN,

I'm David, CEO of Expanso. Today we're launching Expanso Skills — a catalog of 200+ production-ready data processing pipeline recipes for AI agents.

The problem: Every team rebuilds the same data processing primitives from scratch. PII scrubbing, log aggregation, GDPR routing, schema enforcement, dead letter queues. Each time, slightly broken. Each team, from zero.

What we built: Reusable, composable pipeline recipes that run on Expanso Edge (our open-source distributed compute layer). Think npm packages, but for data processing.

A few examples: - `remove-pii` — Strips sensitive fields before data reaches your AI agent - `parse-logs` — 1,000 raw log lines → 1 structured JSON digest (99.9% reduction) - `cross-border-gdpr` — Routes data based on jurisdiction automatically - `dead-letter-queue` — Capture and retry failed pipeline messages - `fan-out-kafka` / `fan-out-s3` — Distribute processed data to multiple destinations - `enforce-schema` — Validate and coerce incoming data before it hits downstream services

Install and run in under 2 minutes: expanso skill install remove-pii expanso skill run remove-pii --input ./customer-data.csv

Why this matters for AI agents: Most agent frameworks assume agents can query whatever they want. That's a security and compliance disaster. Skills enforce least-privilege at the infrastructure level — agents never see raw data, only filtered, purpose-fit outputs.

Product Hunt: https://www.producthunt.com/products/expanso-skills?utm_sour...

Happy to answer anything — architecture, specific skill implementations, edge deployment. Ask away.

Comments URL: https://news.ycombinator.com/item?id=47066006

Points: 2

# Comments: 0

Article URL: https://blogs.microsoft.com/on-the-issues/2026/02/17/acting-with-urgency-to-address-the-growing-ai-divide/

Comments URL: https://news.ycombinator.com/item?id=47065989

Points: 2

# Comments: 0

Article URL: https://gist.github.com/championswimmer/bd0a45f0b1482cb7181d922fd94ab978

Comments URL: https://news.ycombinator.com/item?id=47065983

Points: 2

# Comments: 0

Article URL: https://jvns.ca/blog/2026/02/18/man-pages/

Comments URL: https://news.ycombinator.com/item?id=47065972

Points: 3

# Comments: 0

Article URL: https://electrek.co/2026/02/18/tesla-avoids-30-day-california-sales-suspension-after-dropping-misleading-autopilot-marketing/

Comments URL: https://news.ycombinator.com/item?id=47065971

Points: 2

# Comments: 1

I’ve spent years doing results-based management (RBM) for nonprofits and development programs—mostly designing activities, logic models, and M&E frameworks.

I started by using a custom GPT to improve consistency in early analysis/drafting: https://lnkd.in/dbmBAXMB

Now I’m piloting the next step with OpenClaw, including a focused skill for RBM logic model development: https://lnkd.in/dYuhhN8r

This is still early, but the goal is practical: build specialized agents for development consulting workflows, with human experts making final judgments (not full automation).

Main questions I’m working through: -confidentiality and data handling -security boundaries -governance/accountability of AI-assisted recommendations -how to preserve methodological rigor while speeding up workflow

Long-term, I want to digitize core consulting tasks end-to-end: -problem definition -theory of change -indicators and assumptions -risk logic -adaptation planning

If you work in development programming, M&E, or RBM system design, I’d value feedback and real-world test cases. More technical notes/experiments: https://lnkd.in/d_RwMaFp

Comments URL: https://news.ycombinator.com/item?id=47063762

Points: 1

# Comments: 0

Article URL: https://entropicthoughts.com/solving-systems-of-equations-faster

Comments URL: https://news.ycombinator.com/item?id=47063748

Points: 2

# Comments: 0

Article URL: https://ifp.org/nlm/

Comments URL: https://news.ycombinator.com/item?id=47063747

Points: 1

# Comments: 0

Article URL: https://reclaimthenet.org/arizona-bill-would-require-id-checks-to-use-a-weather-app

Comments URL: https://news.ycombinator.com/item?id=47063724

Points: 2

# Comments: 0

Hi HN

I built a small desktop companion that monitors CLI AI coding agents so you don’t have to stare at the terminal during long tasks.

It shows when the agent is done, needs input, and the current token/context usage (useful to know when it’s about to compact). It’s fully local + free + open source: https://github.com/fredruss/agent-paperclip

It supports Claude Code via hooks and Codex CLI by watching local session files. Default sticker pack is a small Clippy nod (no affiliation, Microsoft please don't sue me).

Next on the to-do list: multi-session visibility.

Would love feedback / issues / stars.

Comments URL: https://news.ycombinator.com/item?id=47063723

Points: 1

# Comments: 0

Article URL: https://www.computerbase.de/artikel/grafikkarten/nativ-vs-dlss-4-5-vs-fsr-upscaling-ai-leser-blindtest-auswertung.96165/

Comments URL: https://news.ycombinator.com/item?id=47063721

Points: 2

# Comments: 0