Engineering
What building Shippy taught us about building agents
The Ai2 team detailed the architecture behind Shippy, a maritime AI agent, which they split into a system-prompt-defined soul, markdown-file…
Engineering
Model Routing Is Simple. Until It Isn’t.
IBM Research argues that routing requests among multiple LLMs in agentic systems is more complex than simple task classification, since…
Engineering
J-Space: Yet Another LLM Mind Reader?
J-Space introduces the Jacobian lens (J-lens), a mechanistic interpretability technique that identifies verbalizable concepts within language model activations. The method…
Engineering
Putting DoctoBERT to Work: A Practical Guide
A practical guide demonstrates fine-tuning DoctoBERT, a French medical language encoder pretrained on the FineMed corpus, across three clinical natural…
Engineering
Deploy GLM-5.2-FP8 as Your Open, Frontier-Level Agent
A guide walks through deploying GLM-5.2-FP8, an open-source frontier-level language model, on a Dell PowerEdge XE9680 server equipped with eight…
Engineering
Giving AI Agents 3D Bodies, Real Jobs, and Wallets on three.ws
Three.ws is a platform that gives AI agents 3D animated bodies, on-chain wallets, and access to paid API tools. It…
Engineering
Verifying Rust Cryptography in SymCrypt, from Standards to Code
Microsoft researchers describe a methodology for formally verifying cryptographic code written in Rust using the Lean proof assistant together with…
Engineering
Flint: A Visualization Language for the AI Era
Flint is an intermediate visualization language that compiles compact, semantic chart specifications into complete configurations for rendering backends including Vega-Lite,…
Engineering
Aurora 1.5: Extending Open Foundation Models for Weather and Earth-System Applications
Aurora 1.5 extends Microsoft’s Earth-system foundation model by expanding its variable coverage from four to twenty-six atmospheric fields, spanning surface,…
Engineering
VKUE: No GPU? Runs Anyway — a 34.7B Reasoner on a Laptop and on Bare CPU
VIDRAFT’s VKUE engine runs a 34.7B-parameter sparse Mixture-of-Experts reasoning model (about 3B active parameters per token) from a single quantized…
Engineering
Programmatic Tool Calling for Agents
The article argues that durable agent reliability comes from workflow design rather than raw model capability gains like GPT-5.6, describing…
Engineering
How to Design an OpenHarness-Style Agent Runtime with Tools, Memory, Permissions, Skills, and Multi-Agent Coordination
This hands-on tutorial builds an OpenHarness-style agent runtime from scratch, reconstructing core building blocks including tool execution, permission governance, lifecycle…