LangSmith sandboxes are now a first-class environment in @harborframework Installation:
pip install "harbor[langsmith]" Authentication:
export LANGSMITH_API_KEY="" That's it. Your evaluations are now ready to run
MACHINE LEARNING
-
LangSmith sandboxes now first-class environment in Harbor
By
–
-
Runway API Launches Recipes for One-Call Generative Media Features
By
–
New on the Runway API: Recipes.
— Runway (@runwayml) 17 juin 2026
Drop production-ready generative media features into your platform, with one API call. Recipes are Runway-built endpoints with our prompting and workflow expertise packaged in. Polished results, without building or maintaining the workflow… pic.twitter.com/mpVolrNL6sNew on the Runway API: Recipes. Drop production-ready generative media features into your platform, with one API call. Recipes are Runway-built endpoints with our prompting and workflow expertise packaged in. Polished results, without building or maintaining the workflow
-
Economic exploitation of unique signals: LangChain Labs study
By
–
We partnered with @FireworksAI_HQ to answer the following question… How can we economically exploit important signals from each unique trace while maintaining state-of-the-art performance? Read our LangChain Labs study
-
Fine-tuning open models can surpass state-of-the-art models
By
–
Fine-tuning open models can surpass or match state-of-the-art models. Base
@Alibaba_Qwen
out of the box with good prompting:
Solid for perceived error classification, below state-of-the-art model performance. With a LoRA SFT job:
Both -
AI model jump felt like Opus 4 to 4.5, GPT-3.5 to 4
By
–
The jump felt to me like Opus 4 to Opus 4.5, or GPT-3.5 to GPT-4
-
Context-aware AI agent Genie Code for data engineers
By
–
For data engineers, data scientists, ML engineers, and analysts, context matters. A useful AI agent needs to understand not just the code, but the data it depends on, the assets it may affect, and the environment where it will run. That’s why we built Genie Code. With Genie
-
SAM 1/2/3 pods discussion: concepts and intent in videogen
By
–
see our SAM 1/2/3 pods with @nikhilaravi and @josephofiowa
, the rise of concepts is definitely part of it, although imo "intent" is more of a videogen problem (see our @EthanHe_42 pod) than pure CV -
New visual identity and website for Llama.cpp
By
–
Llama.cpp has a new visual identity + official website. Run local models today! Now more than ever, open source must prevail. By @alekgrygier and @ggerganov at ggml/hf
-
Distribution of thought paths in a continuous latent space
By
–
Latent Thought Flow This article moves reasoning into a continuous latent space, but instead of learning a single hidden thought path, it learns a distribution over many paths. Using a continuous GFlowNet, Latent Thought Flow assigns a
-
ExpRL uses reference solutions as reward scaffolds for exploratory RL
By
–
“ExpRL: Exploratory RL for LLM Mid-Training” Sparse reward RL works only when the base model can already find useful reasoning paths, but on hard problems it often gets no signal. This paper uses reference solutions as reward scaffolds instead of imitation targets, letting an