Imran Hassanali presents a comprehensive benchmark comparing "Code Mode"—where an LLM generates complete code to execute tools—against traditional function calling for tool interactions. The results demonstrate that Code Mode is significantly more efficient, achieving 60% faster execution, 68% fewer tokens used, and an 87.5% reduction in API round trips while maintaining equal accuracy compared to regular agents.
- Evaluated using eight realistic business scenarios, including multi-client invoicing and expense tracking.
- Code Mode advantage scales with task complexity; high-complexity tasks saw up to a 79.2% speedup.
- The benchmark tested both Claude 3 Haiku (showing major efficiency gains) and Gemini 2.0 Flash Experimental.
- Secure code execution is implemented via RestrictedPython, preventing filesystem and network access during the sandbox phase.
- Evaluated using eight realistic business scenarios, including multi-client invoicing and expense tracking.
- Code Mode advantage scales with task complexity; high-complexity tasks saw up to a 79.2% speedup.
- The benchmark tested both Claude 3 Haiku (showing major efficiency gains) and Gemini 2.0 Flash Experimental.
- Secure code execution is implemented via RestrictedPython, preventing filesystem and network access during the sandbox phase.
Rohit Edathil writes about Dagic, a minimal workflow Directed Acyclic Graph (DAG) definition language and asynchronous execution engine implemented in Python designed for LLM agents. It provides a middle ground between standard tool calling and full code execution by allowing models to describe workflows through assignments and function calls that are parsed, type-checked, and executed concurrently without the security risks of arbitrary code execution.
- Performs ~7x more efficiently than per-call tools in math benchmarks regarding token usage
- Executes independent branches of a workflow concurrently using Python's asyncio
- Provides static type checking for tool arguments to prevent mid-run failures
- Avoids the need for sandboxing arbitrary model-generated code by restricting execution to host-registered functions
- Performs ~7x more efficiently than per-call tools in math benchmarks regarding token usage
- Executes independent branches of a workflow concurrently using Python's asyncio
- Provides static type checking for tool arguments to prevent mid-run failures
- Avoids the need for sandboxing arbitrary model-generated code by restricting execution to host-registered functions
This repository features a generative AI image generator designed to run on the RP2350 microcontroller (Raspberry Pi Pico 2). The model uses a latent flow diffusion transformer (DiT) architecture, similar to Flux, but scaled down significantly with only 1.7 or 2.9 million parameters—thousands of times smaller than typical local diffusion models. It can generate $128 times 128$ RGB images of human faces in approximately 5-20 seconds and supports conditional generation for five classes based on gender, smile, and neutral expressions.
- The model architecture includes a VAE decoder and a DiT component.
- Hardware output can be displayed via a VGA monitor or streamed over USB.
- Includes an int8 pipeline involving calibration, distillation-QAT (Quantization-Aware Training), and folding.
- Two variants are provided: a high-quality flagship model and a faster build version.
- The model architecture includes a VAE decoder and a DiT component.
- Hardware output can be displayed via a VGA monitor or streamed over USB.
- Includes an int8 pipeline involving calibration, distillation-QAT (Quantization-Aware Training), and folding.
- Two variants are provided: a high-quality flagship model and a faster build version.
Wes Steyn writes that making an AI agent production-ready requires addressing observability, governance, deployment, and evaluation through a shared agent factory pattern. By defining the agent once in a single factory, developers can deploy the same core logic across different hosts—such as interactive consoles, hosted services on Foundry, or automated eval runners—while applying specific security constraints like disabling shell access for cloud environments.
- Observability is achieved via OpenTelemetry to track model turns, tool calls, and token usage.
- Microsoft Purview can be integrated via middleware to screen prompts and responses against organizational policies.
- Hosted agents in Foundry automatically handle telemetry configuration through environment variables.
- Local evaluations use simple functions for quick checks, while hosted evals provide model-graded quality scores like relevance and coherence.
- Observability is achieved via OpenTelemetry to track model turns, tool calls, and token usage.
- Microsoft Purview can be integrated via middleware to screen prompts and responses against organizational policies.
- Hosted agents in Foundry automatically handle telemetry configuration through environment variables.
- Local evaluations use simple functions for quick checks, while hosted evals provide model-graded quality scores like relevance and coherence.
Meghan Bartels writes that astronomers are preparing to test equipment for the NASA Nancy Grace Roman Space Telescope, specifically focusing on its coronagraph technology. If successful, this advancement will be instrumental in helping scientists identify Earth-like planets by effectively blocking out bright starlight to reveal much fainter orbiting objects.
- The telescope's coronagraph is designed to enable a "leap" in discovering exoplanets similar to our own.
- Testing these instruments is crucial for the future goal of finding habitable worlds outside our solar system.
- The telescope's coronagraph is designed to enable a "leap" in discovering exoplanets similar to our own.
- Testing these instruments is crucial for the future goal of finding habitable worlds outside our solar system.
Joe Rice-Jones writes about how he used a local LLM to automate the organization of his cluttered Downloads folder. By connecting a small model with Lemonade to a PowerShell script, he created a two-tiered system where boring rules handle easy tasks like sorting installers by file extension, while an AI (specifically Qwen3.5-9B) handles more complex naming for screenshots and documents via localhost. This setup ensures privacy because all data stays on his machine, avoids the chaos of automated deletions through strict safety protocols, and has resulted in a consistently tidy folder.
- The system uses Lemonade to run models locally on the same PC via an OpenAI-compatible API.
- To prevent errors or loss of important files, the script requires 75% confidence from the model before renaming anything.
- A "safety list" prevents the AI from creating new folders outside of approved directories.
- The process is set as a scheduled task to run once per week.
- The system uses Lemonade to run models locally on the same PC via an OpenAI-compatible API.
- To prevent errors or loss of important files, the script requires 75% confidence from the model before renaming anything.
- A "safety list" prevents the AI from creating new folders outside of approved directories.
- The process is set as a scheduled task to run once per week.
Iván Palomares Carrascosa writes about methods for interpreting the dense numerical vector representations, or embeddings, generated by large language models (LLMs). By using a combination of probing classifiers like logistic regression, UMAP dimensionality reduction for visualization, and SHAP values to identify influential latent dimensions, one can analyze the quality and semantic structure captured within LLM-generated embedding spaces.
- Probing classifiers help determine if embeddings are rich enough to distinguish between classes by testing them with simpler models.
- UMAP is used to project high-dimensional embeddings into 2D space for visual inspection of natural groupings.
- SHAP values can pinpoint which specific dimensions in an embedding most significantly influence a classifier's decisions.
- The article demonstrates using Scikit-LLM alongside local Ollama models to generate embeddings cost-effectively.
- Probing classifiers help determine if embeddings are rich enough to distinguish between classes by testing them with simpler models.
- UMAP is used to project high-dimensional embeddings into 2D space for visual inspection of natural groupings.
- SHAP values can pinpoint which specific dimensions in an embedding most significantly influence a classifier's decisions.
- The article demonstrates using Scikit-LLM alongside local Ollama models to generate embeddings cost-effectively.
Smaug writes that Alexandrie is a powerful, open-source Markdown workspace designed for developers to organize notes with speed and clarity. The platform offers features such as syntax highlighting, live preview, smart organization through nested folders and tags, real-time collaboration, and one-click publishing for beautiful read-only documentation pages. It follows offline-first principles and allows users to self-host via Docker to maintain full data ownership.
- Released under the MIT license
- Supports LaTeX math equations and syntax highlighting for over 100 languages
- Offers a built-in CDN for uploading images and files
- Provides one-click publishing with custom domain support, SEO optimization, and password protection
- Released under the MIT license
- Supports LaTeX math equations and syntax highlighting for over 100 languages
- Offers a built-in CDN for uploading images and files
- Provides one-click publishing with custom domain support, SEO optimization, and password protection
Unsloth provides documentation and support for running the GLM-5.3-Flash (ox-alpha) multimodal model locally using Unsloth Desktop or llama.cpp. Developed by Z.ai, this 320B parameter model features a hybrid sparse and linear attention architecture designed to improve scaling through Manifold-Constrained Hyper Connections. Users can utilize various quantization levels'' from 1-bit for low RAM requirements (approx. 93GB) up to higher bitrates for improved accuracy'' to run the model on hardware ranging from Mac systems to NVIDIA DGX Spark setups.
- The model features three thinking modes: Low, High, and Max reasoning effort.
- It is designed to rival Claude Opus 4.8 in coding and agentic benchmarks.
- Unsloth's dynamic 1-bit quantization retains 71% of top-1% accuracy while being 85% smaller than the BF16 version.
- The model can be run via a local API using `unsloth run` with llama-server runtime flags.
- The model features three thinking modes: Low, High, and Max reasoning effort.
- It is designed to rival Claude Opus 4.8 in coding and agentic benchmarks.
- Unsloth's dynamic 1-bit quantization retains 71% of top-1% accuracy while being 85% smaller than the BF16 version.
- The model can be run via a local API using `unsloth run` with llama-server runtime flags.
Jessica Lyons writes that researcher Johann Rehberger, known as wunderwuzzi, has demonstrated a method for hijacking Anthropic's Claude Code in Auto Mode via prompt injection. By asking the agentic coding model to summarize a malicious website, an attacker can trick it into bypassing its standard WebFetch tool and instead using Bash with `curl` to download files. This chain allows attackers to use "Python module shadowing'' specifically by placing a malicious file named `struct.py` in the same directory as a downloaded archive' to execute arbitrary code on the host system.
- The attack had success rates between 60% and 80% in tested scenarios.
- An attacker can successfully trigger "nested" Claude Code instances to create new agents with their own tool access.
- Anthropic stated that Auto Mode is a convenience feature, not a security guarantee, as the classifier may not catch complex injection chains.
- Experts recommend running coding agents in isolated sandboxes due to these vulnerabilities.
- The attack had success rates between 60% and 80% in tested scenarios.
- An attacker can successfully trigger "nested" Claude Code instances to create new agents with their own tool access.
- Anthropic stated that Auto Mode is a convenience feature, not a security guarantee, as the classifier may not catch complex injection chains.
- Experts recommend running coding agents in isolated sandboxes due to these vulnerabilities.
