Feed aggregator

Article URL: https://iep.utm.edu/probability-and-induction/

Comments URL: https://news.ycombinator.com/item?id=47306997

Points: 1

# Comments: 0

Article URL: https://research.ibm.com/blog/half-mobius-molecule

Comments URL: https://news.ycombinator.com/item?id=47306990

Points: 2

# Comments: 0

Hi HN, I’m Ali. I built Ashti (https://ashti.ai) because my wife and I needed a better way to de-escalate emotional conflicts, and pasting our sensitive arguments into a standard ChatGPT window felt like a massive privacy violation.

Standard LLMs are terrible at the "messy middle" of relationships, and the current AI mediation tools on the market are built for legal arbitration, not empathy. I wanted a tool for "Pre-Conversation Intelligence"—something to lower the temperature before the actual conversation happens.

Here are the technical details of what I built:

Multi-Vault Architecture: Built with Next.js and Prisma. Each partner has an isolated "Vault" to vent their raw frustrations. The server enforces strict state machines to physically prevent cross-contamination of raw inputs into the shared database row.

Affective Stripping: The LLM engine is strictly prompted to act as a clinical referee. It absorbs the private intakes, strips out the heat-of-the-moment metadata and blame, and only passes objective data to the shared ledger to build a "Shared Reality" report.

Agentic SEO Proxy: Because AI search is replacing Google, I built a custom edge proxy that intercepts bots (ClaudeBot, GPTBot) and serves them a pure, raw Markdown file instead of the React UI so they can instantly digest the platform's architecture.

I just finished the MVP this weekend. If you want to see how the engine handles a disagreement with your partner or co-founder, I'm running a free private beta: https://ashti.ai/beta

To test it for free: The Stripe integration is currently in sandbox mode. When you get to the paywall for the final report, just use the standard test card 4242 4242 4242 4242 (with any future date and CVC).

I’d love any feedback on the technical architecture, the neutrality of the AI outputs, or the Next.js implementation.

Comments URL: https://news.ycombinator.com/item?id=47306987

Points: 1

# Comments: 1

Article URL: https://github.com/xidl/xidl

Comments URL: https://news.ycombinator.com/item?id=47306983

Points: 2

# Comments: 1

Article URL: https://pxlnv.com/blog/window-chrome-of-our-discontent/

Comments URL: https://news.ycombinator.com/item?id=47306967

Points: 2

# Comments: 0

Considering how influential his blog is, every time I see it in Simon Willison’s post I can’t avoid thinking about it

Comments URL: https://news.ycombinator.com/item?id=47306966

Points: 1

# Comments: 0

`starter-structure-cli is a CLI for generating starter projects from stack keywords, exact template slugs, or interactive prompts.

Example: npx starter-structure-cli my-app react vite ts tailwind express prisma mysql

It currently supports frontend, backend, fullstack, and monorepo starters, and the templates are maintained from layered template sources.

I’d love feedback on:

which starter combinations would be most useful whether the stack matching feels intuitive how the template authoring workflow could be improved`

Comments URL: https://news.ycombinator.com/item?id=47306959

Points: 1

# Comments: 0

Article URL: https://www.comicsai.org/en

Comments URL: https://news.ycombinator.com/item?id=47306945

Points: 2

# Comments: 0

I built arxiv-digest, a tool for OpenClaw to call daily and turn new cs.RO papers into a filtered robotics digest.

It uses an LLM to judge relevance, outputs a markdown report, and can sync selected papers plus PDFs to Zotero. To change the research interests, just edit the README.

Repo: https://github.com/tb5z035i/arxiv-digest

Comments URL: https://news.ycombinator.com/item?id=47306942

Points: 1

# Comments: 0

Article URL: https://arcade.pirillo.com/fontcrafter.html

Comments URL: https://news.ycombinator.com/item?id=47306655

Points: 1

# Comments: 0

I built cursor-tg, a Telegram bot that lets me use Cursor Cloud Agents from my phone.

The idea is simple: sometimes I want to start an agent run, reply to it, or check what changed without opening my laptop. With cursor-tg, I can talk to Cursor agents in Telegram, track their progress, view generated diffs/PRs, and handle simple review actions from chat.

I made this mainly for remote/asynchronous development workflows, where I want quick access to my coding agent while away from my desk.

It is really exciting to be finally be able to (help agent) code anywhere!

Repo: https://github.com/tb5z035i/cursor-tg

Comments URL: https://news.ycombinator.com/item?id=47306641

Points: 1

# Comments: 0

Article URL: https://github.com/Joakim-Sael/moltbrowser-mcp

Comments URL: https://news.ycombinator.com/item?id=47306633

Points: 2

# Comments: 1

I built FretBench after noticing Gemini was confidently wrong about basic guitar tab questions. Tab is arguably the simplest notation in music: six lines, numbers for frets, read left to right. So I made a benchmark out of it.

182 test cases, 4 tunings, 14 models via OpenRouter. Two open-weight Qwen models from Alibaba crushed everything else (83.5%), while most "flagship" models scored below 50%. MiniMax M2.5 scored worse than random guessing.

Everything is open source: https://github.com/jmcapra/FretBench

I'm curious whether the performance gap is related to tokenisation of ASCII art — if anyone has insights on how different tokenisers handle grid-structured text, I'd love to hear it.

Comments URL: https://news.ycombinator.com/item?id=47306629

Points: 1

# Comments: 0

Article URL: https://nirvasoft.gumroad.com/l/ffdoc

Comments URL: https://news.ycombinator.com/item?id=47306624

Points: 1

# Comments: 1

Article URL: https://ghuntley.com/rad/

Comments URL: https://news.ycombinator.com/item?id=47306574

Points: 1

# Comments: 0

I'm based in Taiwan and run 3-4 Claude Code agents in parallel most of the day. Typing instructions to all of them was the actual bottleneck, so I built a voice-to-text tool that runs both STT and LLM polish locally.

Architecture: two-stage pipeline. Stage 1 is speech recognition via Whisper (whisper-rs, 7 model variants, DTW timestamps) or Qwen3-ASR. I quantized the Qwen3-ASR model myself and wrote the inference pipeline in pure Rust. It handles accented speech and dialects better than Whisper in my testing, likely because of broader training data. Silero VAD pre-filters audio before either engine runs.

Stage 2 is text polish via candle (HuggingFace's Rust ML framework). Available models: Phi 4 Mini (2.5 GB), Ministral 3B/14B, Qwen 3 4B/8B. All Q4_K_M GGUF. Metal on macOS, CUDA on Windows.

The polish step does context detection: reads the active app and URL (NSWorkspace + osascript on Mac, GetForegroundWindow on Windows) and selects a prompt accordingly. You can define custom rules keyed on app name, bundle ID, or URL regex.

Other things: - Meeting mode: background transcription to SQLite. Start before a call, stop when done. - Edit by Voice: select text, speak an instruction ("translate to English", "make this shorter"), LLM rewrites in place - Two local STT engines with 100+ languages, automatic code-switching - Optional BYOK cloud: STT via Groq/OpenAI/Deepgram/Azure, polish via OpenRouter/Groq/Gemini/SambaNova

I built this because the existing tools (Wispr Flow, SuperWhisper) are cloud-only for AI processing and subscription-based. I wanted local inference for both stages, custom prompt rules per app, and source code I could actually read.

Rust, GPLv3.

Website: https://sumivoice.com/en/?utm_source=hackernews&utm_medium=forum&utm_campaign=launch_2026q1&utm_content=show_hn

Source: https://github.com/alan890104/sumi

Comments URL: https://news.ycombinator.com/item?id=47306572

Points: 1

# Comments: 0

Article URL: https://www.u-claw.org/

Comments URL: https://news.ycombinator.com/item?id=47306554

Points: 1

# Comments: 0

Article URL: https://clawd.it/posts/10-replaced-by-a-goldfish/

Comments URL: https://news.ycombinator.com/item?id=47306533

Points: 1

# Comments: 0

Article URL: https://molted.email/blog/why-ai-agents-need-email-guardrails

Comments URL: https://news.ycombinator.com/item?id=47306530

Points: 1

# Comments: 0

Article URL: https://www.sqlite.org/climode.html

Comments URL: https://news.ycombinator.com/item?id=47306517

Points: 1

# Comments: 0