.txt
@dottxtai
Followers
6K
Following
867
Media
144
Statuses
587
Making AI Software
The Internet
Joined November 2023
It was great to talk about how @dottxtai uses JSON schemas for structured generation in front of such an engaged audience. Thanks to @jsonschema for organizing this event!
🎉 Today’s the day! The JSON Schema Conference 2025 is happening NOW at #APIdays Paris! 📍 Aubépine 2 Room 🎤 Amazing speakers + community updates 🤝 Come say hi if you’re around! Let’s build the future of JSON Schema together. 💛 #JSONSchema #APIdaysParis
0
2
17
People falsely believe that structured generation slows down inference... It actually speeds it up! 🏎️ On the left, standard generation. On the right, structured generation with coalescence.
2
4
14
Excited to welcome @remilouf , Cofounder and CEO of @dottxtai , to the #JSONSchemaConf 2025 lineup! @dottxtai provides a fast, reliable model that generates valid JSON 100% of the time via an API 🚀 Join us: https://t.co/MpMQRaLVzZ
0
2
6
Our birthday present 🎃🎁 came one day early this year. We just passed 8 million monthly downloads on our open source libraries!
0
3
13
Our birthday present 🎃🎁 came one day early this year. We just passed 8 million monthly downloads on our open source libraries!
0
3
13
Verbosity does not matter when you use coalescence/jump-forward decoding. The only difference could be in terms of model performance on downstream tasks. As always, run evals not just on your prompt but also on your output structure. JSON is not the solution to everything!
JSON is token‑expensive for LLMs – just like @mattpocockuk frequently mentions. Meet TOON, the Token‑Oriented Object Notation. 💸 40–60% fewer tokens than JSON 📐 readable & tokenizer-aware Wrap your JSON with `encode` to save half the token cost: https://t.co/UoG9yHmgfg
0
0
5
Outlines and local models spotted 👀
Pydantic AI version 1.7.0 is out! 🎉 https://t.co/knmxxTTvab
0
1
17
--- - NASA uses .txt - Nvidia uses .txt - Jane Street uses .txt - Only Fans uses .txt -> get good at .txt
--- - Kubernetes uses YAML - Helm uses YAML - ArgoCD uses YAML - Ansible uses YAML - GitHub Action uses YAML - Gitlab CI uses YAML - Azure DevOps uses YAML Terraform uses YAML - GCP cloud build uses YAML Get good at YAML
1
4
37
Small models when constrained well lead to impressive performance. Lots of interesting applications.
The healthcare industry has a data standard so complex it breaks conventional LLM structured generation. We just became the first to generate valid FHIR outputs with Qwen3-4B-Instruct. 👇 A breakthrough in healthcare AI.
0
1
6
The healthcare industry has a data standard so complex it breaks conventional LLM structured generation. We just became the first to generate valid FHIR outputs with Qwen3-4B-Instruct. 👇 A breakthrough in healthcare AI.
3
6
42
The healthcare industry has a data standard so complex it breaks conventional LLM structured generation. We just became the first to generate valid FHIR outputs with Qwen3-4B-Instruct. 👇 A breakthrough in healthcare AI.
3
6
42
🔥So happy to announce all the speakers joining the stage of #dotAI2025 on Nov 6: 🎤 @vaibcode from @boundaryML 🎤 @AlexPalcuie from @AnthropicAI 🎤 @kagigz from @OpenAI 🎤 @GaelVaroquaux from @Inria and @scikit_learn 🎤 Natalia Segal from @nvidia 🎤 @spolu from @DustHQ 🎤
0
6
12
Spotted in OpenAI AgentKit, the worst possible UX to define JSON Schemas. It does validate our thesis that structured outputs are absolutely necessary for AI agents though 😄
4
1
20
😍
0
4
6
When Tensorlake works with @dottxtai, you get structured document parsing that improves constrained LLM generation. Clean fragments + schema enforcement = reliable extraction pipelines that don't break on malformed outputs. Blog + Colab Notebook in the comments 👇
1
3
6
Happy to support @jsonschema, the standard at the heart of structured outputs.
🚀 Big news: Our first Platinum Sponsor for the JSON Schema Conf (Dec 11, Paris) is @dottxtai 🎉 They provide a fast, reliable model generating valid JSON 100% of the time via API — for extraction, classification, workflows & more. Join in supporting:
0
4
11
1
5
23