mcbetz

OpenCode Go, one subscription (currently 10$/month) with multiple models and fair token limits. Sometimes a bit slow, but good value for your buck.

Alternative title: Text to speech in 9.36M, English only.

In Erfurt since 2,5 hours. Out of office train driver keeps us updated from chats with fellow drivers (their sources say it is due to software update), radio is fixed now and trains processed one after another (starting with super fast ones - Munich > Berlin, e.g. - so the tracks get emptied quickly). Other interesting observations: when our train stopped, all hotels were already fully booked, as were coach tickets (Flixbus) that would run in the early morning. Crazy how fast people react to shocks.

Little on differences other than bounding boxes and double the price compared to their previous OCR v3 model from December - https://mistral.ai/news/mistral-ocr-3/ - other benchmarks were used back then.

I observe that music OCR space and the only really good solution so far is soundslice. You scan and review some edge cases and get really good results. Paid service by a small company, very worthy to be supported!

This reminds me of a learning resource that I can't find again: you start with an assessment of how many words you know and then you get new words in context with every session (and maybe some spaces repetition). It was mostly from newspaper articles and catered for every level of English. It was a website (ca 2013), not an app. Any ideas?

I like the direction! But why stop with CRM functionality with very limited schema for tasks on the headless when OP has a lot more to offer there? Why stop at CRM plus Project Management then? Why not also schemas for ERP, HR, finance tools and all other business software? And we even bother with a custom made API when PostgreSQL and PostgREST could reduce the full thinking to the database schema?

Mistral Voxtral has timestamps and diarization and does a good job for German (have not tested for other languages yet).

Have been using it for transactional mails, that was okay. Have not tried it for non-transactional, esp. not in 20+ mio/months.

It's also difficult to find providers for competetive large-scale non-transactional emails, i.e. marketing and newsletter mails.

None comes close to AWS, closest comes are messageflow (PL), elasticemail (PL), brevo (FR). Other players like Scaleway TEM (FR) and Lettermint (NL) don't offer non-transactional.

The video: https://xcancel.com/RuairiRobinson/status/202139494075720913...

You might benefit from LMArena's Leaderboard. It does not have Danish (yet), but German and English evaluation might help: https://lmarena.ai/de/leaderboard/text/german

Openrouter.ai shows the location of providers, you can find just a few European services, but also Singaporean and Canadian. Unfortunately, I could not find a way to filter easily.

Official source for 2025 Q3: 64,1% renewable 20,6% coal 12% gas

https://www.destatis.de/DE/Themen/Branchen-Unternehmen/Energ...

The project was vibe-coded by the author and hence, likely the LLM created that tagline as well. Your service is around a year old (based on https://news.ycombinator.com/item?id=44321099) and LLMs might have picked it up already and also the tagline is straight forward business talk, so no big surprise that multiple people and LLMs come up with the same idea.

Article on the creation: https://www.mikenikles.com/blog/i-vibe-coded-a-database-gui

Author wanted to try vibe coding, despite being a critic. Says that result is unmaintainable.

Any first-hand experiences with the Nxtpaper devices? There's a lot of press coverage of the release official material, especially on the 14 inch model, but few actual reviews. Availability or even infos on availability is an issue, at least in Europe as well.

https://github.com/simonw/llm-cmd supports Anthropic (and a lot more models) to achieve the same.

Documentation of processes with annotations. There's lots of steps, each step with a screenshot and ~10 annotations and if steps change, I need to recreate from scratch. Doable, but also lots of manual work.

Yes, exalidraw or tldraw are drop-in replacements for Figma that make it save and easy to share, but they are general purpose tools, I prefer single-purpose tools with limited features.

I loved Skitch as well and this is a neat app. I wish there was some way of saving a sketch (maybe in LocalStorage) and edit it later (e.g. replacing with new screenshot but keep the markers). I can't remember if Skitch could do it, I use Figma for this today (but it comes with a huge overhead).

Thanks for bringing llmprices.dev to my attention. I have also a comparison page for models hosted on OpenRouter (https://minthemiddle.github.io/openrouter-model-comparison/), I do comparison via regex (so "claude-3-haiku(?!:beta)|flash" will show you haiku, but not haiku-beta vs flash.

I wish that OpenRouter would also expose the amount of output tokens via API as this is also an important criteria.

Does anyone have successfully worked with Non-English text with FTS5 in Sqlite? I could not find any reference for German, e.g. and the default stemming does not seem to work properly (given some short tests).

I used Hackmd in the past to share the draft of my book (1) and liked that people don't need to have an account to comment. Google Docs was no option as no markdown support and account required. The process worked well but I found Hackmd too expensive for just getting feedback. Stash looks promising for this use case.

(1) Written fully in Markdown in Obsidian at this point. I moved to Asciidoc since because of formatting. The early draft is still available on Hackmd though. Details in my bio.

This article compares multiple solutions and recommends docTR (Apache License 2.0): https://source.opennews.org/articles/our-search-best-ocr-too...

Interesting project, wishing you all the best!

If you are using Obsidian, Smart Connections in v2 (1) does also support local embeddings and shows related notes based on semantic similarity.

It's not super great on bi/multi-lingual vaults (DE + EN in my case), but it's improving rapidly and might soon support embedding models that cater for these cases as well.

(1) https://github.com/brianpetro/obsidian-smart-connections

I would add sqlite as well.

I found this detailed comparison of OCRs (both open source and cloud services) super helpful: https://source.opennews.org/articles/our-search-best-ocr-too...

docTR comes out as strongest open solution.

I had good repeated success extracting tables from PDFs using Camelot (Python, https://github.com/camelot-dev/camelot)

Here is a summary of what I read on community building.

- Shared struggles are stronger and last longer than shared interests (That is why a forum for solopreneurs (struggling with so much…) might have a more supportive vibe than a photo community.).

- Be an active and bold moderator: Answer questions quickly and be a role model

- Create clear guidelines for posts (What is allowed? What not?)

- Delete everything that violates the guidelines

- Let users flag posts

- Hunt and ban spammers

- Use captchas for newbie posts

This timeline of the Internet is great and has many sources to dig deeper: https://www.zakon.org/robert/internet/timeline/

You can do this with Tailwind as well: https://tailwindcss.com/docs/extracting-components

// in JS, then build .btn-blue { @apply bg-blue-500 text-white font-bold py-2 px-4 rounded; }