Mick Vleeshouwer
7.9K posts

Mick Vleeshouwer
@iMicknl
Startups & AI @AWSCloud | software engineer & fulltime nerd | Home Automation & @home_assistant contributor | My opinions are my own.
Amsterdam Katılım Ekim 2009
1.8K Takip Edilen1.2K Takipçiler

@levelsio I like that it doesn’t require a cloud subscription and works well with Home Assistant, but it takes quite a while after it rings before I can see the live feed and speak back...
English

@SamuelBeek Gaaf Sam! Ik ben via CharGPT ook steeds meer aan het knutselen, maar loop nu vast. Heb je een invite? 😁

Nederlands

@tomaarsen Sneak peek looked very good, excited for the official publish 😉!
English

Qwen seems to be on the cusp of releasing something very cool, give them a follow here to notice when they go live: huggingface.co/organizations/…

English
Mick Vleeshouwer retweetledi

Happy 50th Birthday, Microsoft! 🎂
To celebrate, we are rolling out GitHub Copilot agent mode to all VS Code users, we’ve added MCP support, we’ve got GAs for Anthropic Claude 3.5, 3.7 Sonnet, 3.7 Sonnet Thinking, Google Gemini 2.0 Flash, and OpenAI o3-mini as well as for Copilot code review and next edit suggestions. And a new Copilot Pro+ tier. ✨
Read more: github.blog/news-insights/…
Satya Nadella@satyanadella
You really know intelligence has been commoditized when CEOs start vibe coding!
English

Will have to do a longer post about the state of VAD tech, but TLDR:
⬆️ much more accurate as it can consider the actual speech meaning when deciding if the utterance is ending
⬇️ slightly slower, hasn't been tested in all customer scenarios
艾略特@elliotchen100
@juberti pros and cons of Semantic VAD?
English

@mitsuhiko Waiting for full @dependabot support… Afterwards I will be migrating all my projects, uv is awesome!
English

@charliermarsh what would be the best way to use Dependabot and UV at the moment? Seems there is no solution for this yet, am I correct? I would love to not move to Renovatebot or other apps/bots.
English

@simonw This was actually my frustration as well, that there were no reliable packages to count image tokens. I started a while ago building 'pictoken' (pypi.org/project/pictok…), perhaps I should finish it some day...
English

If you want to accurately calculate the tokens needed for an OpenAI prompt you need to implement this set of rules for images #calculating-costs" target="_blank" rel="nofollow noopener">platform.openai.com/docs/guides/vi… - and then more-or-less guess what the concatenated JSON for the tool configuration etc looks like when its processed by their API
English

@charliermarsh Why is Dependabot support so behind? What is now the best way to keep track of package update with uv & GitHub?
RenovateBot & other 3rd parties are less preferred…
English

@lucasmeijer Let me know if you figure out how to improve the Dutch ‘accent’ with your system message. I will give it another try tomorrow. 😃
English

Ok, I have the realtime voice api working through azure sweden central. (thanks @iMicknl )
fun first impression: I'm talking to it in dutch, with dutch systemprompt. its response has both audio and transcript. the transcript is in dutch, but the audio in english
English
Mick Vleeshouwer retweetledi

Demo of real-time voice input for a RAG application:
youtube.com/watch?v=vXJka8…
Code @ github.com/Azure-Samples/…

YouTube

English
Mick Vleeshouwer retweetledi

It’s time 🍓. Today, we’re thrilled to open up early access to OpenAI o1- preview and o1-mini for GitHub Copilot Chat in VSCode. And there’s more: it’s also available in our GitHub models catalog, today!
github.blog/news-insights/…
English

@VrijndeSteede Ik dacht precies hetzelfde… vooral omdat er eerst een foto bij stond van een supermarkt (fragment uit video). 🤦♂️
Nederlands

@jeffwilcox Last year Amsterdam banned all motorbikes/mopeds from most bike lanes in the city.
Enjoy your trip!
English

@iMicknl Thanks for the info. I saw the cameras and wasn't sure what they were for. Having motorbikes/mopeds in the bike lanes is something I was told about before the trip and has been surprising in the cities for sure.
English

@LurkerAlpha You can also leverage Document Intelligence on Azure, Lay-out API with markdown output for a similar result.
English

@llama_index Anyway to host on Azure? Required for security reasons.
English

A hidden release that makes LlamaParse way better at RAG over complex documents: we’ve made huge improvements to markdown-based table reconstruction - this allows you to parse very complex tables while making sure the rows/columns are well aligned! 🔨
Check out the results below (source image, before fix, after fix).
Tutorial on how to build advanced RAG over your PDFs with lots of text and complex tables: github.com/run-llama/llam…

English

@SamuelBeek 👍. Die locatie bepaling werkt super om even snel mijn sleutels te vinden, dit zou ideaal zijn voor in m’n Secrid.
Nederlands

@TechPreacher An outlet cover for child safety? Just turn the plug when inserting.
English













