Writer and creator 🚀 Maker of https://xitter.cf/t.co/HJ1qBMXzFW & https://xitter.cf/t.co/Yf4ygA953r & https://xitter.cf/t.co/VZb5taYGT0 🖼️📖⭐ #BuildInPublic .NET & Azure enthusiast
Online 👉
Joined March 2011
- Tweets2.4K
- Following2.6K
- Followers386
- Likes13.7K
Pinned Tweet
My wife’s request for a "together" photo with a friend abroad turned into my latest project. 🛠️
Upload pics of your people (up to 8 for now!), describe the vibe, and let AI bridge the gap.
Live now: pixailer.com
Pavel ✨ keenthinker retweeted
The way to build it also hasn't changed, I think:
1. Ship a thing
2. Show it to people
3. Talk to users
4. Find more users
5. Repeat 🔁
It's just that the product and the competition are very different now.
Pavel ✨ keenthinker retweeted
Can you use EmbeddingGemma 2 for Speech-to-Text?
This experiment uses EmbeddingGemma 2’s unified audio-text vector space to match speech against text via vector similarity.
The result? Transcribing an hour of audio in about 13 seconds, locally!
Pavel ✨ keenthinker retweeted
You can now train your own Decision model with our free notebook! 💡
Qwen3.5-4B will generate decisions instead of text on just 8GB VRAM locally.
Learn to data prep (state, questions, gold answers), train, serve.
Notebook: colab.research.google.com/gi…
Guide: unsloth.ai/docs/basics/train…
You can now train your own Decision model like Jev locally!
We increased Qwen3.5 0.8B’s aggregate accuracy from 20.7% to 74.3% across 3 decision benchmarks - on just 4GB VRAM.
Turn any LLM like Qwen3.8, Gemma 4 into decision models with our open-source Unsloth repo.
We fine-tuned with a Clef head using Unsloth and LoRA (r=64) for one epoch, increasing downstream accuracy from 30–37% to 78%.
GitHub: github.com/unslothai/unsloth
Guide and Notebooks: unsloth.ai/docs/basics/train…
Inspired by a few posts here, I bought this little monster in August. I'm already having a lot of fun with it and the kids. More on that in a few days.
It's amazing how much functionality fits into such a small space.
Pavel ✨ keenthinker retweeted
Mehr Behördengänge bei Adler AI:
🏛️ Steuer-ID erneut mitteilen lassen
🙏 Arbeitsuchend melden, Arbeitslos melden und Arbeitslosengeld beantragen
🫰 Unterhaltsvorschuss beantragen
🔤 Namensänderung. Nachnamen und Vornamen aus wichtigem Grund, der Name nach der Scheidung und der Nachname eines Kindes.
🙆♂️ Verpflichtungserklärung abgeben
🏡 Grundbuchauszug beantragen
♿ Schwerbehindertenausweis beantragen
🙋♂️ Bürgergeld (Grundsicherungsgeld), Grundsicherung im Alter und Hilfe zum Lebensunterhalt beantragen
🛑 Halteverbotszone für den Umzug beantragen
🇩🇪 Einbürgerung und Aufenthaltstitel
🙏 Wohngeld beantragen
Weiter Weiter immer Weiter.
Pavel ✨ keenthinker retweeted
Guten Morgen, Deutschland 🇩🇪
Pünktlich zum Tag der deutschen Einheit unterstützt Adler AI jetzt alle Gemeinden in Deutschland 🎉
Außerdem sind die Behördengänge rund um BAföG, den Hund und Todesfälle nun ebenfalls abgedeckt.
Ein paar Stats an Tag 3:
> 25k Visits seit Start
>2000 Formulare in der DB
~100 Chats pro Stunde
€2500 Kosten bisher
Weiter geht's 💪
Pavel ✨ keenthinker retweeted
.@Cloudflare's decision models are available on Ollama!
Decision models allow you to classify an image, label a bug report or route a support ticket to the right team.
Clef (27B):
ollama pull clef
Clef Flash (9B):
ollama pull clef-flash
🇩🇪/acc 🙂
Ich hab da mal was angefangen.
Aktuell hilft Adler AI mit Infos. Step by Step. Inkl. den nötigen Formularen. So simpel wie möglich.
Ohne Mitarbeit der Behörden wird Adler AI aber nichts *ausführen* oder *einreichen* können. Die Frage ist, ob Step by Step Hilfe für Behördengänge aller Art bereits hilfreich sind.
Weitermachen?
Wow.
Ideogram 4.5 is now live on fal.
Precise edit image model. Make ten edits in a row and 94–99% of everything you didn't ask to change stays identical.
- Typography rendered exactly as written, in any language
- Edit with a mask, or guide changes with up to 4 reference images
- Text to image and editing in one endpoint
Hot take: a single "Dot" from “Dots” is nothing more than “OpenClaw” running on a tiny VPS in the cloud. Or am I wrong? 🤔
I just read a tech article, and one word caught me off guard a little. It was “xeet” - formerly known as “tweet” (“xeeting” - “tweeting”). 😅
So, what are your thoughts - Yes or No?
Very nice addition to the Ollama API (and smart naming too ;). Ollama introduced a new API endpoint "/systemone" for Jev-like queries and models. docs.ollama.com/api/systemon… . Cool. 😎
Oooh, exciting news: "We're happy to share that we've automatically moved your organization from Usage Tier 1 to Usage Tier 2 based on your usage history on our platform.".
I spent 5$ within a week. 😁
🤔💡
why is jev called a "system one" model?
this actually comes from @kahneman_daniel's book "thinking, fast and slow" that outlines two ways of thinking:
system one: fast, cheap, almost automatic (jev)
system two: slow, deliberate, and analytical (llms)
Pavel ✨ keenthinker retweeted
This is *THE* time for tech generalists. Knowing that a thing exists & generally what it's for & not for & then adding a coding agent enables you to build practically anything while still knowing enough to apply guidance & "taste". As @shanselman says, "T"-shaped specialization.
This!
To me, it goes like this: a founder sees another founder be successful. Instead of using this as motivation to find their own journey to build a better product that they truly understand and can build a serious business around, they instead build a copycat or some kind of adjacent product just to benefit from the hype.
Not a big fan of chasing hype for success. It's extremely fleeting and very short-sighted. To me, a founder who tries that, who publicly puts their reputation as a trustworthy and intentional entrepreneur on the line like this, I will think differently about them in the future.
