Hacker News

I was just using ChatGPT to help me pick an SDK library. It mentions a few options by name (e.g. Baileys, whatsapp-web.js), but when I click those names rather than opening a browser with the source page like it used to, it now opens a modal and uses ChatGPT to basically generate a fake homepage for this tool.

From what I can tell, there is no longer any way to easily get to the underlying web page that was referenced in generating its answer to my question.

This feels like a pretty meaningful step backwards. Am I missing something?

Comments URL: https://news.ycombinator.com/item?id=47247891

Points: 2

# Comments: 0

I use Claude every day but there are things I will not type into a cloud service. I have a Mac with Apple Silicon running Apple Foundation Models locally and privately. But I was not always at my Mac. So we built Perspective Intelligence Web. One Mac runs Perspective Server. Any device on your network opens a browser and chats with Apple Intelligence through it. Phone, Windows laptop, Chromebook, Linux machine. Streaming responses, token by token. Nothing leaves your network. MIT License. Next.js, TypeScript, Tailwind. Full writeup: https://taylorarndt.substack.com/p/i-opened-claude-and-then-...

Comments URL: https://news.ycombinator.com/item?id=47247881

Points: 2

# Comments: 0

Three years ago, finding information meant opening Google, clicking 4 links, reading, and piecing together an answer yourself. ChatGPT collapsed that into one step.

We haven't had that moment for taking action yet. To set a recurring water reminder today, you still open your todo app, hit New Task, type it out, set a recurrence, and save. To block focus time, you open Calendar, find a slot, create an event, then update your Slack status separately. Every small action is a small tax. They add up to hours of invisible overhead every week.

GAIA is our attempt at that shift. You tell it what you want done — or it acts before you have to ask.

Gmail & Calendar "Summarise my unread emails and flag anything needing a reply" — runs every morning automatically. "Block 2 hours Thursday for deep work" → calendar event created, Slack status updated. "Draft a follow-up to anyone I haven't heard back from in 5 days" → drafted, waiting for your review.

Slack & Notion "Turn the action items from today's standup into Notion tasks with owners and deadlines". "Remind the team on Slack every Friday at 4pm to update their Linear tickets".

GitHub "Every time a PR is opened, post a summary to the #engineering Slack channel". "Assign new issues labelled bug to me automatically".

Add your own integrations via MCP GAIA has 20+ integrations out of the box. If yours isn't there, you can build your own using MCP, and publish it to the marketplace for others to use. Any existing MCP server works with GAIA out of the box.

The honest version: it works well for workflows you define explicitly, and it's getting better at the ones you don't.

We're a small student team. We've been using GAIA ourselves for quite some time — it genuinely changed how we work, which is why we're releasing it now instead of waiting for perfect.

The design question we keep coming back to: at what point does an assistant acting autonomously feel like it's no longer under your control? We've defaulted to "always reviewable before it sends" — curious how this community thinks about that line.

heygaia.io (no setup) Self-host: docs.heygaia.io/self-hosting GitHub: github.com/theexperiencecompany/gaia

Comments URL: https://news.ycombinator.com/item?id=47247862

Points: 2

# Comments: 2

I wrote a small Zsh plugin that lets you switch macOS Terminal.app profiles directly from the command line.

It's useful if you color-code environments (prod, staging, dev) or when you have a lot of terminals open of the same thing — multiple Claude or Codex sessions, SSH connections, or different projects — and want a quick visual way to tell them apart.

Type `profile` and press Tab to see and autocomplete all available Terminal.app profiles.

Comments URL: https://news.ycombinator.com/item?id=47247852

Points: 1

# Comments: 1

Article URL: https://grith.ai/blog/vibe-coding-killing-open-source

Comments URL: https://news.ycombinator.com/item?id=47247851

Points: 3

# Comments: 0

Article URL: https://danglingpointers.substack.com/p/forest-access-aware-gpu-uvm-management

Comments URL: https://news.ycombinator.com/item?id=47247848

Points: 1

# Comments: 0

Over year ago, our engineering team ditched Scrum. Not because we read some contrarian blog post, but because we looked at our calendars and realized we spent more time talking about work than doing it. Standups that ran long. Grooming sessions nobody prepared for. Retros that concluded "we should have fewer meetings" and then we'd schedule a meeting to discuss that.

We tried going process-free. That lasted a few weeks before leadership started asking "when does this ship?" and nobody had an answer.

So we landed somewhere in the middle. It's called Flights - and it's not new; someone tried it back in 2021: https://simonhoiberg.medium.com/...

The idea is dead simple: a project is a flight. It has a takeoff date, a landing date, and a captain who's responsible for getting it there. Tasks are crates loaded onto the flight. Team members are crew. People doing maintenance between flights are ground mechanics.

That's it. No story points. No velocity charts. No burndown graphs that give everyone a false sense of confidence. A crate is done or it's not. The flight lands on the date or it doesn't.

The thing that surprised us most was how well it communicated upward, sideways and to the rest of the organisation. When we told our CEO "sprint velocity dropped 15%," we'd get blank stares. When we said "this flight is hitting headwinds but still lands Friday," everyone in the room knew exactly what that meant. No translation layer needed.

We started with sticky notes and a shared Figma file. It worked well enough that we built a proper tool around it - an airport-board-style dashboard where you see what's in the air, what's on the runway, and what's landed. Captains, crew, status - all visible at a glance.

We know the obvious reaction: "AI is writing all the code now, who needs project management?" Honestly, we've found the opposite. AI makes individual developers faster, but it doesn't solve the coordination problem. If anything, teams ship more things in parallel now, and the need for visibility into what's actually happening has gone up, not down. Your AI pair programmer doesn't know that another team's flight just went into emergency and yours needs to adjust course.

We've been running it in production for over a year across multiple teams. It was also built as a human-AI collaboration and experiment - Engineers working with Claude from first commit to production, including the database migrations, E2E tests, CI/CD.

Live demo: https://agile.flights Handbook (explains the methodology): https://agile.flights/docs

Happy to answer questions about the methodology and the tool.

Comments URL: https://news.ycombinator.com/item?id=47247408

Points: 1

# Comments: 0

I'm a 52-year-old shift worker at a chemical plant in Finland. Four kids, no coding background. Last Christmas I wanted to send a package to my girlfriend's parents in the Philippines and nobody — not the post office, not DHL, not anyone — could give me a straight answer on price or how to do it. Everyone just said "it's probably expensive." That frustrated me enough to spend my days off building something that answers that question. I used Gemini to write all the code. I copy-pasted. This took about a month. This is what came out: https://github.com/zemloai-ctrl/zemloai-api

Comments URL: https://news.ycombinator.com/item?id=47247405

Points: 1

# Comments: 0

Article URL: https://github.com/danihamm/MontaukOS

Comments URL: https://news.ycombinator.com/item?id=47247392

Points: 1

# Comments: 0

Article URL: https://blog.cloudflare.com/ai-will-transform-programming/

Comments URL: https://news.ycombinator.com/item?id=47247391

Points: 1

# Comments: 0

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

Comments URL: https://news.ycombinator.com/item?id=47247377

Points: 1

# Comments: 0

Article URL: https://aibuilderseries.substack.com/p/sensflationism-the-virus-paralysing

Comments URL: https://news.ycombinator.com/item?id=47247374

Points: 1

# Comments: 0

Article URL: https://en.wikipedia.org/wiki/Lunar_distance_(navigation)

Comments URL: https://news.ycombinator.com/item?id=47247353

Points: 1

# Comments: 0

As agents move from chat to action — invoking tools, delegating to sub-agents, modifying real systems — there's no standard for what a verifiable agent action looks like. Every framework logs differently. Nothing is interoperable. Nothing is cryptographically verifiable. And nobody has defined what "an agent action" even is, formally. We wrote a minimal RFC to fix that: the Agent Action Protocol (AAP). AAP defines:

A canonical AgentAction data model A cryptographic chain integrity model (SHA256-linked, like a git commit graph) Explicit privacy boundaries — raw reasoning and parameters must never be stored, only their hashes An extensibility mechanism for LangChain, OpenAI Agents, Anthropic, etc.

AAP is explicitly NOT a policy engine, a certification body, or a SaaS. It's a primitive. The smallest possible unit of verifiable agent cognition. One design decision we want feedback on: we made abort a first-class decision type, equal to invoke_tool or delegate. An agent refusing to act is as auditable as an agent acting. Does this hold up in your mental model? Reference implementation (Python) + RFC: https://github.com/Thinklanceai/aap-python Tear it apart.

Comments URL: https://news.ycombinator.com/item?id=47247332

Points: 1

# Comments: 0

I am curious about the state of prompt injection attacks on frontier models. Are they still vulnerable? For example, is it safe to let Claude Code look at user-submitted data if it also helps manage some of the infrastructure or code? Can they just be asked to identify prompt injection attacks and flag and ignore them, or do injection attacks change the models' behavior despite the owner's prompts? What are best practices?

Comments URL: https://news.ycombinator.com/item?id=47247328

Points: 1

# Comments: 0

Article URL: https://www.nysenate.gov/legislation/bills/2025/S8102/amendment/A

Comments URL: https://news.ycombinator.com/item?id=47247323

Points: 1

# Comments: 0

I built AIR Blackbox, an open-source static analysis tool that scans Python AI agent code against 6 technical requirements from the EU AI Act (Articles 9, 10, 11, 12, 14, 15). Think of it as a linter for AI governance. To stress-test the scanner — and to see where the industry actually stands — I ran it against 5,754 Python files across 11 major open-source projects. Combined GitHub stars: 341,000+. Projects scanned: AutoGPT (170K stars), Microsoft AutoGen (38K), LlamaIndex (37K), Mem0 (24K), Phidata (18K), LiteLLM (15K), GPT-Researcher (14K), Embedchain (9.2K), LangGraph (8.5K), OpenAI Agents SDK (5.2K), CrewAI Examples (2.8K). Results:

Average compliance score: 2.2 out of 6 articles 97% of files fail Article 9 (Risk Management) 89% fail Article 12 (Record-Keeping) 84% fail Article 14 (Human Oversight) Only 23 out of 5,754 files (0.4%) pass all 6 checks Best scoring repo: AutoGPT at 2.9/6. Worst: CrewAI examples at 1.4/6

What the scanner checks (per article):

Art. 9: risk classification, access control, risk audit Art. 10: input validation, PII handling, data schemas, provenance Art. 11: logging, documentation, type hints Art. 12: structured logging, audit trail, timestamps, log integrity Art. 14: human review, override mechanism, notifications Art. 15: input sanitization, error handling, testing, rate limiting

An article "passes" if at least 1 sub-check is detected. This is generous — real compliance requires substantially more. Caveats I'll save you the trouble of pointing out:

This is static analysis. It can't verify runtime behavior. File-level scanning misses cross-file compliance patterns. The pass threshold is intentionally lenient (1-of-N sub-checks). This checks technical requirements, not legal compliance. It's a linter, not a lawyer.

The EU AI Act enforcement deadline is August 2026. The full report, raw data (JSON), and the scanning scripts are all in the repo.

GitHub: https://github.com/air-blackbox/air-blackbox-mcp Full report: https://github.com/air-blackbox/air-blackbox-mcp/blob/main/b... Install: pip install air-blackbox-mcp Demo: https://huggingface.co/spaces/airblackbox/air-blackbox-scann...

Happy to answer questions about the methodology, the scanner internals, or what we're building next (fine-tuned local LLM for deeper analysis — your code never leaves your machine).

Comments URL: https://news.ycombinator.com/item?id=47247314

Points: 1

# Comments: 1

Article URL: https://jenga.symploke.dev?hn1

Comments URL: https://news.ycombinator.com/item?id=47247312

Points: 2

# Comments: 0

Article URL: https://blog.please-open.it/posts/acr-loa-session-lock-pin/

Comments URL: https://news.ycombinator.com/item?id=47247307

Points: 1

# Comments: 1

Article URL: https://github.com/jundot/omlx

Comments URL: https://news.ycombinator.com/item?id=47247294

Points: 2

# Comments: 1