Hacker News

Article URL: https://www.theguardian.com/world/2026/mar/07/it-means-missile-defence-on-data-centres-drone-strikes-raises-doubts-over-gulf-as-ai-superpower

Comments URL: https://news.ycombinator.com/item?id=47286741

Points: 1

# Comments: 0

Article URL: https://arstechnica.com/space/2026/03/satellite-firm-pauses-imagery-after-revealing-irans-attacks-on-us-bases/

Comments URL: https://news.ycombinator.com/item?id=47286740

Points: 2

# Comments: 0

Article URL: https://rolv.ai

Comments URL: https://news.ycombinator.com/item?id=47286729

Points: 1

# Comments: 1

Article URL: https://github.com/anadim/when-claudes-meet

Comments URL: https://news.ycombinator.com/item?id=47286722

Points: 1

# Comments: 0

Hi HN,

We built Shadow, a real-time meeting assistant that listens to your call and surfaces useful prompts while you’re speaking.

The idea came from a simple observation: during important conversations (sales calls, founder meetings, interviews) you often realize the best question or insight 5 minutes too late.

Shadow tries to solve that by acting like a quiet wingman during the call.

What it does right now: • Tracks agenda topics so conversations don’t drift • Suggests follow-up questions based on what’s being said • Surfaces useful facts or angles in real time

This is still an early MVP and we’re trying to figure out: 1. What type of conversations this helps most with 2. Whether prompts should be more proactive or subtle 3. How much assistance is actually useful without becoming distracting

Would love feedback from the community. Product: https://shadowlabs.ai

Comments URL: https://news.ycombinator.com/item?id=47286702

Points: 1

# Comments: 0

Article URL: https://cartalk-amp.exe.xyz/

Comments URL: https://news.ycombinator.com/item?id=47286701

Points: 1

# Comments: 0

Article URL: https://alexdanco.com/2021/01/22/the-michael-scott-theory-of-social-class/

Comments URL: https://news.ycombinator.com/item?id=47286679

Points: 1

# Comments: 0

Article URL: https://uwplse.org/2026/02/24/egglog-containers.html

Comments URL: https://news.ycombinator.com/item?id=47286661

Points: 1

# Comments: 0

Most CLI players have very rudimentary visualizations, so I focused on creating a "Quantum Spectrum Analyzer"—a high-frame-rate, physics-driven rainbow visualizer that uses gravity-based peak falling and mirrored symmetry.

Under the hood, it manages a pool of ffplay instances and uses OS-level process suspension (via ctypes on Windows and signal on POSIX) to provide an "instant-off" pause experience without cutting the audio buffer mid-stream.

Check it out here: https://github.com/iamekabir-web/Nirvana

Comments URL: https://news.ycombinator.com/item?id=47286659

Points: 3

# Comments: 0

Article URL: https://www.ribbonfarm.com/2009/10/07/the-gervais-principle-or-the-office-according-to-the-office/

Comments URL: https://news.ycombinator.com/item?id=47286657

Points: 1

# Comments: 0

I got tired of AI code review tools that just run a single LLM pass over a diff and call it a review. So I built something different.

Open Code Review spins up multiple AI reviewers from your CLI, each with a different focus. They review independently, then debate each other’s findings, then synthesize into a single review. You get a local dashboard that spins up out of your repo to navigate everything.

The debate step is what makes it work. A single-pass reviewer hallucinates issues and misses real ones. When agents argue with each other, the false positives get killed and the real problems get surfaced with better reasoning. My team’s senior engineers have been consistently surprised by what it catches.

Latest release adds GitHub integration — post the synthesized review directly to PRs. You can post the raw synthesis or have it ghostwritten in your voice. Your call on whether that’s ethical lol.

Runs locally from your CLI, not a SaaS. No catch, pure open source. Works on any repo, any language. Fully open source. The dashboard serves from your project directory, not a separate app.

We’ve been using this daily on our own team for months. It’s the only AI review tool that actually stuck in our workflow.

Repo: https://github.com/spencermarx/open-code-review

Would love feedback from anyone who tries it, especially on the multi-agent synthesis quality. That’s the part I’m most interested in improving.

Comments URL: https://news.ycombinator.com/item?id=47286654

Points: 1

# Comments: 0

I was wasting a lot of time on X and after uninstalling it from my phone, I noticed how anxious it was making me. But I still missed knowing what’s going on with AI, news, and the people I follow. So I built Somree. It summarizes Twitter/X accounts and YouTube playlists into a clean daily email digest. You pick who and what you want to follow, and it does the rest — no feed, no algorithm, no rabbit holes. How it works: ∙ Add Twitter accounts and YouTube playlists you care about ∙ Somree fetches and summarizes the content using AI ∙ You get a daily email with the highlights Tech stack: SvelteKit, Convex, Gemini 2.5 Flash for summarization, Resend for email delivery. It’s $6/month. Would love feedback from anyone who tries it.

Comments URL: https://news.ycombinator.com/item?id=47286650

Points: 2

# Comments: 0

I'm a full-stack developer (Python backend + Flutter mobile/web) working in the AI/LLM space. I'm working at Beijing China, And my day job pays about $2,500/month. I've built AI systems for companies in finance and healthcare — RAG pipelines, knowledge bases, AI agents, the whole stack. I can ship a complete product solo, from backend to deployment.

Last year I decided to seriously try making money on the side. Here's what happened:

*Attempt 1: Built a product (Telegram subscription bot)* Built a content subscription platform on Telegram — paid tiers, built-in shop, automated operations. At first it kinda worked: ~$70/month on ads, ~$110-140/month revenue. Tiny margin, but hey, it's positive. So I tried to scale — pushed ad spend to ~$280/month. Revenue? Only ~$220. Scaling literally made it worse. Turns out the small profit was just a lucky pocket of cheap traffic, not a real business. Shut it down.

*Attempt 2: Partnered with someone to build a product* Found a partner — I build, they sell. Built the whole thing. They tried promoting it for about two weeks, didn't see quick money, and just... stopped. Product died on the vine.

*Attempt 3: Another partnership* Different person, different project. I did all the dev work, delivered the finished product. Partner ghosted. Never heard from them again.

*Attempt 4: "Trust me, it's easy money"* Someone convinced me their business model was printing money. I invested ~$800 to get in. Lost all of it. Classic lesson, expensive tuition.

*Attempt 5: Freelancing* Started looking at freelance projects. Reality check: a $7,000 project takes one person roughly 2 months of full-time work. As a side gig on top of a day job, that's brutal. The per-hour math is honestly depressing once you factor in communication, revisions, and scope creep.

*Attempt 6: Small gigs* Just did a simple website for a friend — $300 for 2 days of work. Fine as a favor, but obviously not a path to meaningful income.

*Net result after one year: -$2,200 and mass of time gone.*

Here's what I think I've learned (but I'm not sure I've learned the right lessons):

- I can build things, but building ≠ earning - I keep falling into the "trade time for small money" trap - Partnerships have burned me every single time — I do the work, the other side flakes - I don't know how to find clients who actually pay well - My product attempt failed because I had no real distribution strategy — just threw ads at it and hoped - I'm apparently an easy target for "easy money" pitches

I'm not looking for "learn to code" advice — I can code fine. What I can't figure out is the business/money side.

For those who've broken out of the "skilled but broke" cycle:

- What actually changed things for you? - Did you find a specific niche? Change how you find clients? Build a different kind of product? - Is there something obvious I'm missing?

Genuinely asking. I'm not trying to promote anything — I just want to understand what I'm doing wrong.

Comments URL: https://news.ycombinator.com/item?id=47286643

Points: 2

# Comments: 1

Article URL: https://www.carlos-menezes.com/posts/ai-sentiment-march-2026

Comments URL: https://news.ycombinator.com/item?id=47286629

Points: 1

# Comments: 0

Article URL: https://www.france24.com/en/humanitarian-crisis-looming-300-000-displaced-as-israeli-army-pounds-lebanon

Comments URL: https://news.ycombinator.com/item?id=47286617

Points: 2

# Comments: 0

Article URL: https://tomtunguz.com/qwen-9b-matches-frontier-models/

Comments URL: https://news.ycombinator.com/item?id=47286611

Points: 1

# Comments: 0

A browser based mushroom identification tool running the Swin vision transformer model. It can classify thousands fungal species in just 200ms.

All processing happens locally in your browser. Images don't leave your machine.

Comments URL: https://news.ycombinator.com/item?id=47286590

Points: 1

# Comments: 0

Article URL: https://www.economist.com/finance-and-economics/2026/03/05/americans-electricity-bills-are-up-dont-blame-ai

Comments URL: https://news.ycombinator.com/item?id=47286582

Points: 2

# Comments: 0

Article URL: https://www.warmto.me

Comments URL: https://news.ycombinator.com/item?id=47286576

Points: 1

# Comments: 1

Article URL: https://research.swtch.com/acme

Comments URL: https://news.ycombinator.com/item?id=47286539

Points: 1

# Comments: 0