Comments URL: https://news.ycombinator.com/item?id=48523664

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48523534

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48522598

Points: 3

# Comments: 0

I'm kindly asking for your participation in the open beta for my AI-managed mini-games website. Thank you in advance!

For a limited time window, I'm setting the all-free feature flag to true. I hope you have a lot of fun exploring the AI's sense for games! Here and there, I tweaked it to help with visual consistency.

I would be deeply grateful if you opted into analytics.

$2,300 in API tokens...

Cheers!

Comments URL: https://news.ycombinator.com/item?id=48522555

Points: 45

# Comments: 51

Comments URL: https://news.ycombinator.com/item?id=48522417

Points: 11

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=48522082

Points: 4

# Comments: 2

you launch a product and share on social media ,and expect your icp will try it .

But among those ,there are people will malicious intent trry to test you system and want to create several accounts using disposable emails to try to get more free stuff without paying

i kow its hard ,emailverify does that

stop spam signups,

https://emailverify.se

Comments URL: https://news.ycombinator.com/item?id=48522075

Points: 2

# Comments: 2

Comments URL: https://news.ycombinator.com/item?id=48521901

Points: 2

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=48521848

Points: 6

# Comments: 3

Comments URL: https://news.ycombinator.com/item?id=48521682

Points: 2

# Comments: 0

Here are some of the cool features of our platform:

Explore Mode: instead of specifying the steps of a user journey yourself just give an agent a sentence or two describing what it looks like and the agent figures out the steps itself so it can follow them in the future.

A/B Testing: You can tweak system prompts for the web agent and A/B test it within the platform itself.

Self-healing: the steps for a journey update over time to align with your product's evolution. So ideally over time the amount of time and the number of tokens the agent takes trend down as it needs to make fewer decisions and think less.

Comments URL: https://news.ycombinator.com/item?id=48521381

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48521284

Points: 2

# Comments: 0

I'd been using MacWhisper for meeting notes, which is great and actually very powerful, but before every call I had to set it up, and half the time I forgot and walked out of an hour-long meeting with nothing written down. I wanted what I was using MacWhisper for, but launchable in one shortcut that just starts in the background and then leaves me alone.

You press a global shortcut and a small bar shows up at the bottom of the screen (there's another shortcut to hide it if you'd rather not see it working), and it records your mic and the system audio. The system side is captured with a Core Audio process tap, so no bot joins the meeting and it works with any app, or just a conversation in the same room.

Transcription runs on-device, with two engines to choose from. A fast one (Parakeet-TDT) and a slower, more accurate one (Whisper large-v3-turbo via WhisperKit), both through Core ML on the Neural Engine. The first time you transcribe it pulls the models from Hugging Face, and after that it works fully offline. Audio never leaves the Mac, it's sandboxed, there's no account and it collects no telemetry.

As I was building it I wanted to bake in a couple of workflows I'd wished for in other transcription apps.

1. Mid-meeting you can press another global shortcut to mark a "key moment" and type a quick optional note. It shows up inline in the transcript at that timestamp. I kept catching myself thinking "wait, that bit matters" in meetings and reaching to jot it down, which for me defeated the point of having notes taken automatically. I use it all the time. When I paste the transcript into an LLM afterwards, the important moments are already flagged so it doesn't gloss over them, which is more noticeable in longer meetings with lots of topics. 2. While it's recording there's a rough live recap so you can glance at what's been said so far, again behind a global shortcut.

Each session is just a folder on disk containing mic.wav, system.wav, transcript.json and transcript.md. The system track gets speaker labels (Speaker 1, Speaker 2, and so on) and your mic track is labelled Microphone. You can't rename the detected speakers yet, but I'm working on it.

It won't summarise for you (you get a clean transcript plus the key-moment markers, then bring your own LLM), and it's Apple Silicon only, because the models run on the Neural Engine.

There's also an optional Google Calendar connection that reads your events to auto-name sessions.

If you've used Granola, Otter or Fathom, the difference is they put a bot in your call or do the work in the cloud, sometimes both, and I wanted neither.

I built Trace for myself to fix a pain point I had with existing tools, and it's done that for me, so I wanted to share in case it helps someone else. Open to thoughts, roasting, or any other feedback.

It's £9.99 on the App Store.

Comments URL: https://news.ycombinator.com/item?id=48521236

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48520871

Points: 2

# Comments: 1

Headless mode allows you to use the AI as a command-line utility for automation and scripting. In Claude Code you run it with the -p flag: claude -p, in codex - exec, opencode - run.

2. Ask human

The traditional communication channel with the operator won't work in headless mode - we need to implement a dedicated tool. Here is an example of how this can be done https://github.com/sermakarevich/claude/tree/main/mcp/ask_hu...

3. Tasks queue

Beads is a lightweight distributed graph issue tracker for AI agents, powered by Dolt. You can create tasks, define dependencies between tasks, and have status, priorities, hierarchy. Beads helps prevent multiple tasks from being claimed by > 1 worker.

4. Worker artifacts

We want to be able to monitor how a worker is doing, at what stage it is, and resume it after a restart. For every task we can create a dedicated folder using the beads task id and put into it what we need. I put there: - plan and status md - knowledge md - events.jsonl - stderr

The worker is instructed in its prompt to check if artifacts exist, which allows it to proceed from where the job was left.

5. Worker isolation

To prepare to run multiple workers we need to isolate them. Git worktree can be used here. I am testing this approach: - worker gets the task and implements it - the next worker, spawned automatically, validates the task is done, tests it, merges the worktree, closes the ticket and creates another one for a fix if required

6. Multiple workers

To be able to run multiple workers we need a simple orchestrator. An infinite loop constantly checking beads / config and triggering new workers when required.

7. Coder agnostic

A worker can be basically any coder. I started with Claude, added Codex and Agy. And last added Opencode.

8. Subscription limits.

3 coding agents can burn the Claude $200 subscription limit in 30 minutes even if you switch to Sonnet 4.6. API tokens cost x40 compared to tokens in the subscription - this is too expensive. The idea I am testing is: - use the strongest model possible to analyse/design and add tasks - use a local model as a worker - use a stronger model to validate workers and add new tasks to fix potential misimplementations

I am using the qwen3.6:36B local model with Ollama, deployed on 2 GPU cards, 36GB in total, with a 256K context window. This is slower, but it is free of charge. And surprisingly it worked, and worked way better than I would expect it to. Fable 5 was extremely great at creating clear and simple tickets until it was.

Another approach I was considering is Bedrock qwen, paying per token, or renting a 96GB GPU for $1400 per month.

I found that it's optimal to run 3 workers concurrently even though Ollama processes 1 request at a time. The reason is the ask_human tool. If a worker asks me something at night - it has to wait until morning doing nothing. Running three +/- guarantees GPU load at 100%.

9. Nice integrations

UI - to observe tasks / beads / config / chat / analytics

It's easy to miss when a model asks a question. It's visible in the UI - a green circle near chat, but that's it. So I added a Telegram integration - now I receive questions from workers on Telegram and can reply there, get the status of tasks, create new tasks etc.

I am doing this for my PoC projects ofc: - improving fleet - building a data collection and analysis related app

What I am seeing is that 24x7 coders are closer than I thought they are. Even weaker models can deliver good results when the task is simple and well defined. All components for building these systems are there.

Repo: https://github.com/sermakarevich/fleet

Comments URL: https://news.ycombinator.com/item?id=48520757

Points: 4

# Comments: 1

The problem I kept running into is that when I use an LLM for research, it gives me citations that look right but aren't most of the time. Sometimes the paper doesn't exist. Sometimes it does exist but it's been retracted. Most times the sources are not fully vetted nor truly relevant to my search. I got tired of checking these manually.

That's why I built this updated MCP. It looks the work up in Crossref to confirm it's real. It checks the Retraction Watch data to see if the paper was pulled. It also does a basic match between the claim I'm making and what the source actually says, so I know which references to read before I trust them.

You can also hand it a whole bibliography in BibTeX, RIS or CSL-JSON and it tells you which entries don't resolve or have been retracted. Dead links get saved to the Wayback Machine so the reference still points somewhere later.

It's a single Go binary. Open Source (MIT) and runs against most respected search engines including specialized domain-specific ones.

If you have the same problem, I hope this is useful to you too. Appreciate your feedback.

Comments URL: https://news.ycombinator.com/item?id=48520752

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48520719

Points: 3

# Comments: 0

Therefore, I decided to create local-memory-mcp, a local MCP that allows LLMs to read and write to a RAG. It also has built in systems that help keep the system efficient like a max chunk size and deprecation when information gets updated.

I would love it if you all would try it out and let me know your thoughts. It's been a huge time and token saver for us.

Thanks for the help and enjoy!

Comments URL: https://news.ycombinator.com/item?id=48520532

Points: 3

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=48520502

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48520360

Points: 5

# Comments: 0