Hacker News

Article URL: https://buttondown.com/jaffray/archive/expectation-and-copysets/

Comments URL: https://news.ycombinator.com/item?id=46950690

Points: 1

# Comments: 0

Hi HN, we built PaperPod: agent-native sandboxes for code execution, live URLs, browser automation, LLM inference, persistent memory, with 50+ built-in tools (ffmpeg, sqlite, etc.) - all via our CLI tool or HTTP endpoints.

The problem: Existing infra tools require SDK setup, API keys, or clicking through UIs, things agents can't do well.

What we built: On-demand sandboxes accessible via our CLI or HTTP endpoints. Try it right now with any coding agent:

npm install -g @paperpod/cli ppod login your@email.com ppod help

What's included: - Code execution (Python, JS, shell) - Live preview URLs - Persistent memory (10MB, survives restarts) - LLM Inference - Browser automation (screenshots, PDFs, scraping via Playwright) - 50+ tools (ffmpeg, sqlite, git, pandoc, imagemagick)

$5 free credits on signup (~14 hours), no credit card required. Built on Cloudflare Containers.

Quick Demo with OpenClaw: https://youtube.com/shorts/gTbyz26mPxk?feature=share

What would make this more useful for your agent workflows?

Comments URL: https://news.ycombinator.com/item?id=46950654

Points: 1

# Comments: 1

Article URL: https://ksaweryskowron.substack.com/p/cycle-time-is-not-a-number

Comments URL: https://news.ycombinator.com/item?id=46950646

Points: 1

# Comments: 0

Article URL: https://www.businessinsider.com/instagram-meta-working-on-internal-prototype-disappearing-messages-snapchat-2026-2

Comments URL: https://news.ycombinator.com/item?id=46950028

Points: 1

# Comments: 0

Article URL: https://briefings.canaryiq.com/p/the-invisible-governor

Comments URL: https://news.ycombinator.com/item?id=46950011

Points: 1

# Comments: 0

Article URL: https://dnedic.github.io/blog/firmware-linker-script-generation/

Comments URL: https://news.ycombinator.com/item?id=46950010

Points: 1

# Comments: 0

Hey HN! We're Roanak and Andrew from Trainy (YC S23). We build GPU infrastructure for ML teams (scheduling multi-node training jobs on Kubernetes). When Neptune announced they're shutting down, our customers didn't have a clear path forward. The alternatives exist but none of them matched the UI experience Neptune had, especially at scale. So we decided to build Pluto.

Pluto is an open-source experiment tracker based on our fork of MLOp. The main idea is that you can add one import alongside your existing Neptune code and it logs to both platforms simultaneously. You validate that everything matches on real training runs, then when you're ready, set an env var and all Neptune API calls redirect to Pluto. We also built a Neptune exporter for historical runs.

We're focusing heavily on having a UI that stays responsive at scale, since ML teams can have thousands of runs per project and the tracker is open all day. If you find anything slow or buggy in the playground, we'd love to hear about it.

Beyond the compatibility layer, we're working on: tensor logging with on-the-fly visualization (log raw tensors instead of pre-rendered plots), code diffing between runs, and Linear/Jira integration. We just shipped Pluto MCP in alpha, which lets you query experiment data with an LLM.

Live playground (no signup): https://demo.pluto.trainy.ai/o/dev-org/projects/my-ml-projec...

Quickstart: https://docs.trainy.ai/pluto/quickstart

Listed on Neptune's official transition hub: https://docs.neptune.ai/transition_hub/migration/to_pluto

Let us know what you think! We'd especially love feedback from anyone managing experiment tracking across a team.

Comments URL: https://news.ycombinator.com/item?id=46950009

Points: 1

# Comments: 0

Article URL: https://www.bloomberg.com/news/articles/2026-02-09/alphabet-mandates-banks-for-rare-100-year-sterling-bond

Comments URL: https://news.ycombinator.com/item?id=46950003

Points: 1

# Comments: 0

Article URL: https://zachwills.net/building-at-the-speed-of-thought/

Comments URL: https://news.ycombinator.com/item?id=46949978

Points: 1

# Comments: 0

Article URL: https://www.tyla.com/news/super-bowl-ring-doorbell-ad-privacy-concerns-099156-20260209

Comments URL: https://news.ycombinator.com/item?id=46949969

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/watch?v=163MUrEoj5Y

Comments URL: https://news.ycombinator.com/item?id=46949957

Points: 1

# Comments: 0

Article URL: https://github.com/youngkent/clawtime

Comments URL: https://news.ycombinator.com/item?id=46949945

Points: 1

# Comments: 0

Article URL: https://sdfprotocol.org/

Comments URL: https://news.ycombinator.com/item?id=46949915

Points: 1

# Comments: 1

I built llm-use, an open-source tool to run AI agent workflows across multiple LLMs with routing and cost optimization.

Repo: https://github.com/llm-use/llm-use

OpenClaw-style agents are powerful but get expensive if every step runs on a single high-end model. llm-use helps by: • using a strong model only for planning and final synthesis • running most steps on cheaper or local models • mixing local and cloud models in the same workflow

Example:

python3 cli.py exec \ --orchestrator anthropic:claude-4-5-sonnet \ --worker ollama:llama3.1:8b \ --task "Monitor sources and produce a daily summary"

This setup keeps long-running agents predictable in cost while preserving quality where it matters.

Feedback welcome.

Comments URL: https://news.ycombinator.com/item?id=46949890

Points: 1

# Comments: 0

Article URL: https://www.nature.com/articles/43360

Comments URL: https://news.ycombinator.com/item?id=46949879

Points: 2

# Comments: 0

Article URL: https://capnfabs.net/posts/foamcore-would-be-a-sick-name-for-a-music-genre/

Comments URL: https://news.ycombinator.com/item?id=46949870

Points: 2

# Comments: 1

Article URL: https://www.youtube.com/watch?v=Bc0bIpDVEa8

Comments URL: https://news.ycombinator.com/item?id=46949864

Points: 2

# Comments: 0

Article URL: https://moldandyeast.substack.com/p/fat-agents-vs-solver-markets

Comments URL: https://news.ycombinator.com/item?id=46949842

Points: 1

# Comments: 1

I work at a PSP. We recently had a production bug where a payment succeeded but our system didn't update — the webhook was delayed 45 seconds. Stripe's test mode doesn't simulate delays, timeouts, or webhook failures. We ended up writing internal mocks, but it's tedious. Considering building a simple tool: configurable delays, webhook failure simulation, request timeline. Curious how others handle this. Do you just write custom mocks? Accept the risk? Use something I don't know about?

Comments URL: https://news.ycombinator.com/item?id=46949828

Points: 1

# Comments: 3

Article URL: https://ntietz.com/blog/using-an-engineering-notebook/

Comments URL: https://news.ycombinator.com/item?id=46949814

Points: 1

# Comments: 0