Hacker News

I think this is apropos of the "Show HN" tag, as the post is explanatory and the entire codebase this little side-story use case discussed in TFA is in the repo and free to use. (I'd be pleased if you did!)

In the post, as I tried to capture in the title submitted, I outline my journey of exploration, when I became determined to make GitHub-Flavored Markdown display my text, with color, style and alignment of my choosing, which as I discovered after setting out to do so, the inability to do such a thing outside of fenced blocks with pre-defined syntax highlighting is a well-known condition, which is met with "works as intended" response because, well, GitHub doesn't want their repos looking like MySpace or Geocities or presenting security risk exposure by allowing arbitrary html/CSS styling. Sure, I should have used GitHub Pages to build a page from my Markdown using Jekyll, which is a supported way to control the styling of your own documents in your repo, but where's the fun in that?

The linked post documents the workaround I arrived at, which became an output target format that nobody has ever asked for from my ASCII line-Art diagramming tool. I thought some here might appreciate the documentation of "wasting my time so you don't have to" on a technical solution for a problem I probably just shouldn't have cared about and moved on.

Comments URL: https://news.ycombinator.com/item?id=47203584

Points: 1

# Comments: 0

Article URL: https://andrewkchan.dev/posts/diffusion.html

Comments URL: https://news.ycombinator.com/item?id=47203573

Points: 1

# Comments: 0

Hey HN, I built a free AI flashcard generator with no signup and no paywall. Quizlet charges $7.99/month, Revisely caps free users at 3 generations, Knowt keeps moving features behind paywalls. I wanted to build the opposite of that. https://prepareyourself.app Paste text or upload a PDF. Pick an output type — flashcards, quiz, summary, or explain-with-examples. Get results in ~10 seconds. Done. I've been running it for 28 days. Here are the real numbers:

- 180 users. 8 of them engaged deeply. One person hit "review again" 64 times on a single flashcard deck. The other 172 bounced.

- Indian users spend 2x longer than US users (96 sec vs 44 sec). Bengaluru and Ahmedabad are my strongest cities. No idea why.

- Flashcards get 10x the engagement of quizzes. Quiz completions in a month: 1 person. So I made flashcards the default.

- Day 7 retention: 0%. Literally zero. Just shipped spaced repetition reminders — no data yet on whether it helps.

- Revenue: $0.

So the product works for the few who try it. The problem is getting people to try it.

Features: 4 output types (flashcards, quiz, summary, explain-with-examples), PDF upload, TTS audio on every card, spaced repetition, export to PDF/DOCX/JSON, 11 languages. History and streaks stored client-side in IndexedDB — nothing leaves the browser unless you hit generate.

Stack: Next.js, Supabase, Claude Haiku, Vercel. No auth required — rate limiting via IP + fingerprinting.

Would love feedback on UX, output quality, or whether "free forever, no signup" is a viable positioning against VC-funded competitors who are all moving toward paywalls.

thank you!

Comments URL: https://news.ycombinator.com/item?id=47203567

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/watch?v=BYXbuik3dgA

Comments URL: https://news.ycombinator.com/item?id=47203551

Points: 1

# Comments: 0

Article URL: https://www.newscientist.com/article/2517389-human-brain-cells-on-a-chip-learned-to-play-doom-in-a-week/

Comments URL: https://news.ycombinator.com/item?id=47203544

Points: 2

# Comments: 0

Article URL: https://www.atlasobscura.com/places/malm-whale

Comments URL: https://news.ycombinator.com/item?id=47203533

Points: 1

# Comments: 0

Article URL: https://en.wikipedia.org/wiki/Plugtest

Comments URL: https://news.ycombinator.com/item?id=47203527

Points: 1

# Comments: 0

Your agent doesn't know your codebase. It knows the code. Not the decisions behind it.

Why you picked that queue over a direct DB write. Why that module has no external dependencies. Why the retry logic is in the client and not the server. That context lives in your head. The moment the session ends, it's gone.

EmCogni Code fixes this. One Go binary. Local MCP server. You annotate decisions as you make them:

#example 1 emcogni note payments --type decision "Async via queue — direct writes caused timeout spikes under load"

#example 2 emcogni note payments --type invariant "Never call external APIs inline — always enqueue"

Any MCP-compatible agent (Claude, Cursor, Copilot) gets that context automatically before touching a file, via get_context tool. EmCogni can also propose context additions via propose_context tool.

Free for local/solo use at emcogni.com.

Team/managed tier coming — shared context across engineers and agents, org-level knowledge base. Waitlist open at the same link.

Curious how others handle this. CLAUDE.md files? Custom RAG? Just re-explaining every session?

Comments URL: https://news.ycombinator.com/item?id=47203522

Points: 1

# Comments: 0

Article URL: https://felixbarbalet.com/simple-made-inevitable-the-economics-of-language-choice-in-the-llm-era/

Comments URL: https://news.ycombinator.com/item?id=47203508

Points: 1

# Comments: 0

Article URL: https://en.wikipedia.org/wiki/Idiot_plot

Comments URL: https://news.ycombinator.com/item?id=47203496

Points: 1

# Comments: 0

Article URL: https://gvanrossum.github.io/interviews/Thomas.html

Comments URL: https://news.ycombinator.com/item?id=47203486

Points: 3

# Comments: 0

Article URL: https://translatorhub.org/

Comments URL: https://news.ycombinator.com/item?id=47203460

Points: 1

# Comments: 0

I've been using Claude Code alot and keep running into the same issues: losing track of which session is doing what, sending Claude off on a task only to discover 30 minutes later it had immediately asked a clarifying question and was sitting there waiting, lack of easy remote access, and just a plethora of worktree issues (worktree off branch, auto-installing dependencies, etc).

So I built ClaudeTerminal — an Electron app that is inspired by Windows Terminal but is purpose-built for Claude Code. It ain't fancy - it feels like Terminal, but it does have:

Tabbed sessions — Run multiple Claude Code instances side by side, each with status icons showing if they're working, idle, or need input

Desktop notifications — Get notified when a background session finishes a task or needs your attention. Click the notification to jump straight to that tab

Remote access — Access your sessions from any device via a temporary Cloudflare tunnel. Check on your agents from your phone

Session persistence — Close the app, reopen it, and every session resumes exactly where it left off

Git worktree integration — Ctrl+W creates a new worktree and opens a Claude session scoped to it (branches from your current branch, not just main!). Worktree names show on the tabs.

Tab auto-naming — Your first prompt gets analyzed by Haiku (via call to claude call terminal) to generate a descriptive tab name (so you see "Auth Bug Fix" instead of "Tab 3")

Repository hooks — Trigger scripts on lifecycle events (e.g., auto-run pnpm install when a worktree is created)

Shell tabs — Mix in PowerShell/WSL tabs alongside your Claude sessions

Keyboard-driven — Ctrl+T for new tab, Ctrl+W for worktree tab, Ctrl+Tab to switch, Ctrl+1-9 to jump, etc.

It's open source (MIT): https://github.com/Mr8BitHK/claude-terminal.

Windows is the primary platform (yup, I don't use a Mac - sorry all!), but I've included experimental macOS/Linux builds.

Would love feedback from other Claude Code users and hope this helps... someone :)

Disclosure: I'm the sole developer. ClaudeTerminal is free and open source (MIT). No paid tier, no telemetry, no sign-up. Image shamelessly AI generated. Tool co-developed with Claude Code.

Comments URL: https://news.ycombinator.com/item?id=47203452

Points: 1

# Comments: 0

Article URL: https://tanelp.github.io/neurips2021/

Comments URL: https://news.ycombinator.com/item?id=47203436

Points: 1

# Comments: 0

Article URL: https://en.wikipedia.org/wiki/Office_of_Technology_Assessment

Comments URL: https://news.ycombinator.com/item?id=47203427

Points: 1

# Comments: 0

Article URL: https://ostechnix.com/midnightbsd-excludes-california-digital-age-assurance-act/

Comments URL: https://news.ycombinator.com/item?id=47203410

Points: 4

# Comments: 2

Article URL: https://github.com/alibaba/OpenSandbox

Comments URL: https://news.ycombinator.com/item?id=47203397

Points: 1

# Comments: 0

When a current system (even with hardware-level MTE memory tag expansion) successfully intercepts a Buffer Overflow attack, what is its instinctive reaction? It immediately terminates the program, generates a Crash Log in the background, and might even pop up a notification telling the user, "Malicious attack blocked."

This isn't defense; it's the system acting as a free real-time debugger for APT (Advanced Persistent Threat) groups.

In Project Genesis's underlying security architecture, we will completely redefine "defense." Security mechanisms shouldn't be about showcasing how hard the system is working; they should be about trapping attackers in an endless war of attrition.

1. *Refusing to Report Errors: The Silent Drop in the Face of Zero-Day Vulnerabilities*

We must acknowledge that no system can predictably patch all vulnerabilities; patch gaps always exist. When a nation-state cyber army or top hackers use a legitimate app as a Trojan to attempt to overstep the limits by writing to the L1 cache or tampering with system metrics, the Genesis system's underlying layers (powered by 30% resident SNN-driven security gatekeepers) physically intercept the attack and will absolutely not provide any error code.

*Intelligence Deprivation:* Malicious packets are directly dropped into the "silent drop." The feedback received by the Trojan program remotely is not "Access Denied," but rather "Timeout" or a highly realistic "natural crash."

*Ward Off Attitude:* Hackers have no way of knowing whether their payload failed due to network instability, a code error, or a genuine MTE alert. This tactic of depriving users of debugging information will significantly lengthen the development cost of the attack chain, driving their shift engineers to the brink of collapse under endless blind testing.

2. Rejecting Complacency: Combating Low-Level Rogue Software with a "Public Review Mechanism"

For hackers who exploit vulnerabilities to launch high-level attacks, we remain oblivious; but for the rampant rogue apps that excessively demand permissions, we must strike hard.

Covering up the truth will only lull users into a false sense of security. Genesis introduces a *"Medium/Low-Level Risk Public Review Mechanism":* During routine Play Store or system scans, the system will no longer simply report "safe." When big data analysis reveals that a photo editing app abnormally wakes the microphone 50 times a day in the background, the system will directly mark it as a "Medium Risk" in the health report.

Risk Transfer and Responsibility Decentralization: The UI will not forcibly delete it, but instead provide a one-click "Remove Unnecessary Permissions" downgrade button. The system lays bare the harsh truth: "You can keep it, but at your own risk."

The "flame effect" acts as a censorship weapon: when a large number of users see the app's "medium-level vulnerability" warning on a red background, screenshots inevitably circulate on social media. This public relations disaster and mass uninstallation from the market forces those unscrupulous developers to fix it overnight within 24 hours, which is more effective than any official API restriction order.

3. The 30-Day Immune Purge

Vulnerabilities are often discovered after the fact. To prevent users from unintentionally opening permission backdoors, the Genesis system has a built-in 30-day cycle of forced cleanup.

Based on local SNN frequency monitoring and anonymous comparison of cloud big data, the system automatically downgrades all "over-authorized and infrequently used" API channels to the minimum survival standard every month. Even if a hacker manages to implant a backdoor into a legitimate app, the backdoor's lifespan is physically limited to 30 days, completely severing the root of a long-term botnet.

In summary: True security is about playing dumb against high-level threats and cracking down on low-level malicious actors.

Comments URL: https://news.ycombinator.com/item?id=47203396

Points: 1

# Comments: 0

Article URL: https://www.bloomberg.com/news/articles/2026-02-28/polymarket-iran-bets-hit-529-million-as-new-wallets-draw-notice

Comments URL: https://news.ycombinator.com/item?id=47203386

Points: 2

# Comments: 0

Hey HN, Most AI “agents” I’ve tried are basically chatbots with amnesia — they forget everything the moment you close the tab and can’t do anything unless you’re sitting there watching them. I wanted real AI coworkers that just… work. So I built Computer Agents (aiOS). Every agent you create gets its own isolated computer in the cloud — complete with persistent memory, a real file system, code execution environment (with automatic dependency management), and the ability to run scheduled or webhook-triggered tasks 24/7. You give it a goal (“research this market and email me a report every Monday”, “generate floor plans from client briefs”, “handle incoming support emails”, “run my weekly data analysis”), walk away, and come back to finished results in your inbox, Telegram, or dashboard. Key highlights: • Persistent workspaces — context and files survive forever (no more “remember what we talked about last week?”) • Native iOS app (iPhone + iPad) + native Mac app + web dashboard • Python + TypeScript SDKs (pip install computer-agents, npm install computer-agents) • Multi-agent orchestration (sequential, parallel, map-reduce, conditional flows) • Built-in skills: deep web research with citations, web search, image generation, full code interpreter • Integrations: Email, Telegram, GitHub, Google Drive, OneDrive, Notion, webhooks, etc. • Runs in secure isolated cloud containers (you own your data) It’s live at https://computer-agents.com Free tier gives you 150 compute tokens (~15–23 decent-sized tasks) so you can try it right now. Pro starts at $19/mo when you want more. This is very much still a young indie project (I’m the solo founder), but it’s already helping real teams automate support, research, content, and coding workflows. Would love your honest feedback — especially: • What persistent/long-running agent pain points have you hit with other tools? • Interesting use cases you’d want to try? • Thoughts on the architecture (sandboxing, persistence model, orchestration) Happy to answer any questions! Thanks, Jan Luca (indie maker behind Computer Agents) P.S. If you’re into computer-use agents, we also have a comparison page: https://computer-agents.com/compare/computer-use-agents

Comments URL: https://news.ycombinator.com/item?id=47203385

Points: 3

# Comments: 0