Words fail me.

The amount of tokens spent to infer the day of the week lol

www.data-monger.com/syndeblog/th...

Couldn't agree with this more!

The easy part is that it's clearly a very nice way to start a conversation with someone.

A secondary effect (which is a bit hard to quantify) is that doing things like this helps other people to know what things interest you, which can be a source of opportunities.

Need to extract a few highlights from a news article?

Try this out. We used the article about Neural Magic's recent acquisition by Red Hat.

You can even change your structure to easily add reasoning and observations for better output quality.

Check the gist out: https://buff.ly/4hhCFVz

Happy building this week, everyone! Let us know what you're excited about working on.

ChatOutlines | 🦜️🔗 LangChain This will help you getting started with Outlines chat models. For detailed documentation of all ChatOutlines features and configurations head to the API reference.

Did you know you can use Outlines with @langchain.bsky.social?

🚨 NEW BLOG POST ALERT 🚨

We wrote a case study about .gifter, a demonstration of a simple LLM-powered web applications with structured generation (Outlines) and Flask!

It's also a tool to help you buy gifts for your loved ones, in case you need ideas.

Link: https://buff.ly/4fenrz5

It's a new year, so we wanted to check in and see what types of content we make is the most useful.

Would you like to see more
- Blog posts (less code, more think pieces)
- Videos (deep dives + short overviews)
- Cookbooks (focused code)
- Demos (bigger projects)

alright pfam what AI engineering resources should we make

What's the biggest problem you face in building AI applications?

Ai Builders Mastering AI Agents, LLMs, RAG, and Reasoning

I have TWO big presentations on January 23rd! I will be busy.

One is LIVE in San Francisco with @dottxtai.bsky.social (me),
@neo4j.bsky.social, Modal, and Neural Magic.

The other is VIRTUAL with
@odsc.bsky.social!

ODSC: www.summit.ai?utm_campaign...

OSS AI: lu.ma/2jacrv79

Most agent frameworks are (ironically) low-agency tools, because they take control + internals away from engineers.

Granted, they can provide some powerful abstractions.

But these abstractions are too early -- we don't know how to build AI systems. We should build from first principles.

Structured generation tools like Outlines are AI engineering primitives -- most agents should be using structured generation by default rather than as an afterthought.

How to use AI to generate gift ideas | Use a robot to become a mildly better gift giver Description.gifter helps you find thoughtful gift ideas by analyzing descriptions of the recipient. It uses large language models to generate customized gift...

Are you a terrible gift giver? Would you like to use LLMs to become a slightly better gift giver?

Check out our new video!

Vid: https://buff.ly/3ZLYnep

@cameron.pfiffer.org shows off .gifter, a language model-powered gift recommendation engine. Features support for web search with Exa.

This is actually good to show skeptics of structured generation @dottxtai.bsky.social. Demonstrates very clearly that you can stop the model from fighting the constraints by telling it how it’s constrained, even if those constraints might seem absurd (like only using words from the King James Bible)

Guix Social virtual meet-up - Last one of 2024!, Thu, Dec 12, 2024, 6:00 PM | Meetup **Guix Social with talk by Christine Lemmer-Webber** Christine will be talking about the work the Spritely Institute has been doing on bringing Guile Scheme to WASM ([Hoot!

Heya y'all, I'll be speaking at Guix Social this Thursday about Spritely! www.meetup.com/guix-social/...

I'll have a talk, but a lot of it I'm also interested in having be audience conversation about Guix, Spritely, Guile, Scheme, decentralized networks, and what we see as the future together!

🏮 DEMO ALERT🏮

Introducing the Robot Santa Game, a terminal-based game that can be played by you or a language model.

This showcases how to permit a model to choose from a set of actions reliably, as is often the case with games.

Check out the demo code: https://buff.ly/49uryWK

I didn't even realize this was something we could vote on!

YouTube: https://www.youtube.com/watch?v=csw6TVfzBcw

Code: https://github.com/dottxt-ai/demos/tree/main/logs

🏮 New YouTube video! 🏮

We experimented with a log monitoring system. Spin up the agent and it'll monitor your logs for any potential issues -- it works with webserver logs like nginx or Apache, Linux system logs, etc.

This is an important distinction -- I recommend reading this if you use some kind of tool to structure output.

Creating a Structured AI Log Analysis System with Python & LLMs YouTube video by dottxt

🏮 New YouTube video! 🏮

We experimented with a log monitoring system. Spin up the agent and it'll monitor your logs for any potential issues -- it works with webserver logs like nginx or Apache, Linux system logs, etc.

youtu.be/csw6TVfzBcw

Link to a recording of my talk: www.youtube.com/live/x94JRyz...

Understanding Context-Free Grammars: An Interview with .txt's Daniel Quernheim! YouTube video by dottxt

New video! @cameron.pfiffer.org talks to our computational linguist Daniel Quernheim.

Learn what context free grammars are, and how to make JSON where every string has to start with the letter A.

youtu.be/rMHdpo_D9m0?...

Check out our new YouTube video. We're trying out kind of an interview format, so let me know if you love/hate this

lesson: if you care about the performance of something, you gotta run your own benchmarks

People on Bluesky seem to have more nuanced technical takes about structured gen and language modeling.

We love it.

The "Let Me Speak Freely" got way too much traction too easily, despite a few serious issues.

Well aware of Brandolini's law we published a thorough rebuttal of the main conclusion, highlighted the key methodological flaws and taught a few things in the process.

Hope you enjoy it!

This is an excellent rebuttal to a paper that I think gained way too much traction too easily. I really appreciate that the .txt team took the time and effort to put together such a well made case in favor of guided generation.