Feed aggregator

Article URL: https://twitter.com/MikeIsaac/status/2100691714713509986

Comments URL: https://news.ycombinator.com/item?id=49759034

Points: 1

# Comments: 0

Article URL: https://thenewstack.io/claude-code-parallel-projects/

Comments URL: https://news.ycombinator.com/item?id=49759033

Points: 1

# Comments: 0

Article URL: https://geastack.com/blog-same-app-color-and-e-paper

Comments URL: https://news.ycombinator.com/item?id=49759020

Points: 1

# Comments: 0

I built onPanda – an interactive tool for token visualization & control, model inspection and data annotation.

onPanda is designed for hackers, power users, curious minds, and engineers. Its UI is built for deep exploration and efficient data annotation.

- The core loop is simple: hover over a token → click an alternative or edit freely → continue generation. You can edit every part of model output exposed by onPanda, including reasoning and tool calls. - Edit prompts directly, branch tool calls, and use a tree structure to record branch history. This makes onPanda useful for model inspection and prompt engineering. - Support multiple modalities, including images, video, and audio; use tool calls and connect MCP servers to perform tasks in real environments. - Connect popular harnesses such as Claude Code, Codex, and OpenCode to execute tasks. Explore and compare their tool sets, system prompts, skills, and memory mechanisms. - onPanda includes a built-in browser-agent, an agent that runs in the user's browser without installation. It uses the browser as its harness and provides JavaScript execution, information retrieval, interface interaction, multimedia I/O, local file access, and persistent memory. - onPanda stands for on-Policy Alignment Data Annotator. - As an annotation tool, onPanda efficiently labels on-policy data. Its token-level correction format also provides fine-grained supervision with precise positions and naturally paired positive–negative samples. - We believe these unique properties will make token-level correction a highly efficient and practical paradigm for future LLM alignment. - Our paper and benchmark for token-level correction: https://on-panda.github.io/research/

I have been building onPanda since Sept 2024, and it took two years for it to gradually enrich its functionality and ease of use. Any feedback and evaluation are welcome.

Try it online (works on mobile): https://onpanda.diyer22.com/

GitHub repo for self-hosting: https://github.com/on-panda/on-panda

Detailed introduction on X: https://x.com/diyerxx/status/2101020850405462041

Comments URL: https://news.ycombinator.com/item?id=49759013

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/watch?v=k0o1p4plC98

Comments URL: https://news.ycombinator.com/item?id=49759009

Points: 1

# Comments: 0

Article URL: https://techcrunch.com/2026/09/17/base-labs-launches-an-open-weight-ai-safety-partnership-with-hugging-face-and-goodfire/

Comments URL: https://news.ycombinator.com/item?id=49758977

Points: 2

# Comments: 0

Article URL: https://techxplore.com/news/2026-09-advertising-ai-chatbots.html

Comments URL: https://news.ycombinator.com/item?id=49758974

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/watch?v=_j806JHhCRo

Comments URL: https://news.ycombinator.com/item?id=49758970

Points: 2

# Comments: 0

Article URL: https://www.techradar.com/pro/whatever-you-try-to-fix-breaks-things-even-more-how-to-avoid-the-hidden-dangers-of-ai-code

Comments URL: https://news.ycombinator.com/item?id=49758946

Points: 1

# Comments: 0

Article URL: https://www.reuters.com/world/anthropic-quietly-sets-up-biology-lab-it-ramps-ai-drug-program-2026-09-18/

Comments URL: https://news.ycombinator.com/item?id=49758943

Points: 1

# Comments: 1

Article URL: https://debunkbot.com/

Comments URL: https://news.ycombinator.com/item?id=49758937

Points: 2

# Comments: 0

Article URL: https://gitlab.com/allddd/go-systemd-time

Comments URL: https://news.ycombinator.com/item?id=49758918

Points: 1

# Comments: 0

hey hn, i've been building peelaway, a tool for creating and editing images that starts with a curated gallery of visual styles.

the idea is pretty simple: pick a style you like, then make something with a photo, a prompt, or both. you can turn a photo of your cat into a screen print, or try a paper-craft or miniature look.

Peelaway started with image editing, and over time i moved the style gallery to the center of the product. i often heard users describing how hard it is to describe a style, so im interested in how much of the process can start with seeing something you like, instead of figuring out how to describe it.

there’s also an api and mcp server if you want to use the editing tools from code or an agent.

would love to hear what you try, where it falls short, and whether picking a style is a useful starting point for you.

peelaway.io

Comments URL: https://news.ycombinator.com/item?id=49758906

Points: 1

# Comments: 0

Article URL: https://day3.app/resend-pricing-calculator

Comments URL: https://news.ycombinator.com/item?id=49758901

Points: 1

# Comments: 0

Article URL: https://yello.sh

Comments URL: https://news.ycombinator.com/item?id=49758895

Points: 1

# Comments: 1

Article URL: https://twitter.com/trycua/status/2101014004927729737

Comments URL: https://news.ycombinator.com/item?id=49758885

Points: 1

# Comments: 0

I was frustrated by having 10 terminal windows open, and I checked a few solutions on how to make this better. They are all nice, but then I realized what I really wanted was something that feels a bit deeper and more intelligent than a "dashboard" of all your agents.

For example, I routinely have 3+ agents exploring different frontiers of the same idea. How do I keep them inline? What if I see 3 more promising directions from one of them? How do I tie up all the loose ends? If I see one agent got a brilliant observation, how do I easily spread that to other agents?

So I started exploring many ways to connect agents together. That led me to realize that the actual fundamental things are not the agents, but the substance. Understandings, decisions, insights, etc. Can I quickly understand what's said? Can I quickly react to it? Can I quickly find it two days later? Can I quickly share to other agents? When ideas and insights are scattered across agents, can I organize those? Can I easily share with the next new agent?

I tried to make optimizations through those angles. Though still far away from my vision, there is enough value that I think it's interesting to share. A few concrete features:

- Agents proactively leverage graphs and annotations to explain ideas

- You can comment inline on text, docs, html files. React to things right where they are

- You can find messages or docs from days, weeks ago by describing what you remembered instead of using exact keywords. Agents can too.

- You can organize topics and sessions in hierarchical or graphical structure (automatically). Or ask an agent to do it if you want it in some particular flavor. In those session maps, everything remains clickable and interactive, with real time status too. So you can have an agent to start 10 sessions in the morning, draw their dependencies and relationships, and watch each agent work through their task until everything in the whole map is finished. Add, remove, tweak, or reorganize the tasks as you wish.

Anyway, that's a lot of words... I don't expect you to read all that. If you find it interesting enough, give it a try!

Comments URL: https://news.ycombinator.com/item?id=49758880

Points: 1

# Comments: 0

Article URL: https://captain-memo.ispcq.com/

Comments URL: https://news.ycombinator.com/item?id=49758858

Points: 1

# Comments: 0

Article URL: https://github.com/kaarelkaarelson/lingbot-world-v2-realtime

Comments URL: https://news.ycombinator.com/item?id=49758853

Points: 2

# Comments: 1