Feed aggregator

Hi HN — I’m the founder of ChexHQ.

We’re building a financial decision intelligence platform for finance teams that are tired of spreadsheet-heavy forecasting and manual consolidation.

The goal: help finance leaders get real-time visibility into runway, cash flow, and key metrics without building complex models from scratch.

We’re early and would love feedback — especially from CFOs, finance managers, or operators here.

What would make something like this genuinely useful for you?

Comments URL: https://news.ycombinator.com/item?id=47013187

Points: 1

# Comments: 0

Article URL: https://www.helionenergy.com/articles/helion-achieves-new-fusion-energy-milestones/

Comments URL: https://news.ycombinator.com/item?id=47013136

Points: 1

# Comments: 0

Article URL: https://pydissem.rgoswami.me/

Comments URL: https://news.ycombinator.com/item?id=47013134

Points: 1

# Comments: 0

Hi! I built Cute Magick because I wanted the simplicity & power of early web dev, but not the fragility.

It's an open source, self-hostable web host where your sites are plain folders of real code — HTML, PHP, Python, Node, Lua, SQL — and every file save creates a snapshot you can preview and rewind in one click.

No build steps, no abstractions. (Literally -- these are per-request runtimes).

Source: https://github.com/pinkhairs/cutemagick

Site: https://cutemagick.com

Docs: https://pixelswithin.notion.site/Cute-Magick-Docs-2fdb91326d...

Comments URL: https://news.ycombinator.com/item?id=47013131

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=47013110

Points: 1

# Comments: 0

Article URL: https://www.nytimes.com/2026/02/14/business/china-chips-nvidia-huawei.html

Comments URL: https://news.ycombinator.com/item?id=47013101

Points: 1

# Comments: 2

My father thinks in Bengali, but typing in native script is painful. Virtual keyboards are slow, punctuation is a maze, and switching between languages breaks flow. For formal writing — emails, letters, posts — it was exhausting.

So I built him a tool where he could just speak.

He dictated a two-page letter in Bengali, read the transcription, and said: "I didn't even need to change a single comma."

That's when I knew this needed to exist.

Parrot transcribes and translates speech in 11+ Indian languages. The difference: it's significantly more accurate than Whisper on Hindi, Tamil, Telugu, Bengali, and handles code-mixed speech (Hinglish, etc.) naturally. Outputs in native scripts (নমস্কার not "nomoshkar") or Romanized.

Sits in your menu bar. Press a hotkey, speak in any app — WhatsApp Web, Word, Slack, browsers. It writes.

Built for everyone who thinks in their language but struggles with virtual keyboards that don't correct grammar or formatting.

Desktop app for Windows & macOS. Free to try.

Shipping improvements weekly. Would love feedback from anyone dealing with multilingual input.

Comments URL: https://news.ycombinator.com/item?id=47013088

Points: 1

# Comments: 0

Article URL: https://techxplore.com/news/2026-02-flattery-debate-ai-mirror-human.html

Comments URL: https://news.ycombinator.com/item?id=47013073

Points: 1

# Comments: 0

Hi HN — I built cgrep, a local-first, code-aware search tool for AI coding agents (and humans).

The goal is to reduce noisy retrieval loops and token waste in real repositories. cgrep combines BM25 + tree-sitter symbol awareness, with optional semantic/hybrid search, and returns deterministic JSON for agent workflows.

What it does: - Code navigation: definition, references, callers, dependents - Focused context tools: read, map - Agent flow: `agent locate` -> `agent expand` (small payload first, expand only selected IDs) - MCP support: `cgrep mcp serve` + host install helpers - Agent install support: claude-code, codex, copilot, cursor, opencode

Benchmark snapshot (PyTorch, 6 implementation-tracing scenarios): - Baseline (`grep`) tokens-to-complete: 127,665 - cgrep (`agent locate/expand`) tokens-to-complete: 6,153 - 95.2% fewer tokens (20.75x smaller) - Avg retrieval latency to completion: 1321.3ms -> 22.7ms (~58.2x faster after indexing)

Links: - Repo: https://github.com/meghendra6/cgrep - Docs: https://meghendra6.github.io/cgrep/ - Benchmark method/results: https://meghendra6.github.io/cgrep/benchmarks/pytorch-agent-...

I’d really appreciate feedback on: - Real-world agent workflows I should benchmark next - MCP/agent integrations I should add - Cases where cgrep retrieval quality still falls short

Comments URL: https://news.ycombinator.com/item?id=47013067

Points: 2

# Comments: 0

Article URL: https://infosec.exchange/@mttaggart/116065340523529645

Comments URL: https://news.ycombinator.com/item?id=47013059

Points: 3

# Comments: 0

Article URL: https://www.openfirmware.info/

Comments URL: https://news.ycombinator.com/item?id=47013042

Points: 1

# Comments: 0

Article URL: https://cybrkyd.com/post/showcasing-my-git-repositories-on-the-web/

Comments URL: https://news.ycombinator.com/item?id=47013040

Points: 1

# Comments: 0

Article URL: https://github.com/edelprino/DCIEM

Comments URL: https://news.ycombinator.com/item?id=47013030

Points: 1

# Comments: 0

Article URL: https://www.scientificamerican.com/article/mathematicians-launch-first-proof-a-first-of-its-kind-math-exam-for-ai/

Comments URL: https://news.ycombinator.com/item?id=47013001

Points: 1

# Comments: 1

Article URL: https://hrml.dev

Comments URL: https://news.ycombinator.com/item?id=47012998

Points: 1

# Comments: 1

Hey HN! I recently open-sourced a project that I came up with in the late 2022 and have been working on and off since. It's taken around a year or so of active development in total.

My problem with habit-trackers is that they're all the same. All of them basically allow to track only one thing - X times per 1 day. That's very rigid and goes against my instinct to ease into new things.

I wanted something that's flexible, self-sustainable, and wouldn't require a lot of maintenance.

After a bit of tinkering I came up with something novel - adjustable habit frequencies. And made all the custom github/anki-style heatmaps with all that functionality on top, meaning stuff like heatmaps for once-in-three-days habits.

It's hard to put into words, as it's very visual, so here's the landing: https://neohabit.org/ (might be a bit slow as it's deployed with github-pages)

And a demo: https://neohabit.org/projects

GitHub: https://github.com/Vsein/Neohabit

Comments URL: https://news.ycombinator.com/item?id=47012984

Points: 1

# Comments: 0

Author here. Built this after working on AI agent security at Radware, where we discovered ZombieAgent - persistent malicious instructions in agent memory.

The insight: Don't teach agents to resist attacks. Virtualize their perceived reality so attacks never enter their world. Like VMs hiding physical RAM → agents shouldn't see raw dangerous inputs.

ARCHITECTURE: - Input virtualization: Strip attacks at boundary (not after agent sees them) - Provenance tracking: Prevents contaminated learning (critical with continuous learning coming in 1-2 years per Amodei) - Taint propagation: Deterministic physics laws prevent data exfiltration - No LLM in critical path: Fully deterministic, testable

Working PoC demonstrates: - Prompt injection prevention (attacks stripped at virtualization boundary) - Taint containment (untrusted data can't escape system) - Deterministic decisions (same input = same output, always)

CRITICAL TIMING: Dario Amodei (Anthropic CEO, Feb 13): Continuous learning in 1-2 years [1] Problem: Memory poisoning + continuous learning = permanent compromise Solution: Provenance tracking prevents untrusted data from entering learning loop

Research context: - OpenAI: "unlikely to ever be fully solved" [2] - Anthropic: 1% ASR = "meaningful risk" - Academic research: 90-100% bypass rates on published defenses [3]

Seeking feedback on whether ontological security (does X exist?) beats permission security (can agent do X?) for agent systems.

Practical workarounds available in repo for immediate use while PoC matures.

Disclaimer: Personal project, not Radware-endorsed. References to published work only.

Happy to answer questions!

[1] https://www.dwarkesh.com/p/dario-amodei-2 [2] https://simonwillison.net/2024/Dec/9/openai-prompt-injection... [3] https://arxiv.org/abs/2310.12815

Comments URL: https://news.ycombinator.com/item?id=47012965

Points: 1

# Comments: 0

Article URL: https://github.com/PulseBeat02/yt-media-storage

Comments URL: https://news.ycombinator.com/item?id=47012964

Points: 7

# Comments: 6

Article URL: https://brandonli.net/spectra/doc/

Comments URL: https://news.ycombinator.com/item?id=47012895

Points: 1

# Comments: 0