Tech News
GLM-5.3 is now open-weight
GLM-5.3, an open-weight model, is released, with developers praising its performance and ease of use compared to alternatives like DeepSeek Flash.
Introducing Hy4 Preview
New open weight text input (no vision) LLM from Chinese company Tencent today: 770B total parameters, 49B active parameters, 1M token context window, 1.56TB on Hugging Face.
Build agentic creative workflows with Amazon Quick and fal
Creative teams produce more assets than ever, but fragmented tools and manual context transfer slow production.
Our decision on Cursor following its acquisition by SpaceX
OpenAI bans Cursor from its services after the AI code editor was acquired by SpaceX, citing ToS violations similar to Anthropic's earlier ban on xAI.
I accidentally turned LLM memory into program analysis
A developer shares how they repurposed LLM memory into a program analysis tool, sparking discussion on using LLMs only at the terminals of request fulfillment with mechanical reasoning in between.
GitHub Repos
koala73/worldmonitor
Real-time global intelligence dashboard.
# TypeScript# agent# aiGraphify-Labs/graphify
Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph.
# Python# ai agents# antigravitytrpc-group/trpc-agent-go
A Go framework for building production agent systems with graph workflows, tools, memory, A2A, AG-UI, MCP, evaluation, and observability.
# Go# a2a# a2a protocolResearch Papers
CritICL: Inference-Time Weak-to-Strong Generalization from Small Language Model Failure Modes
CritICL is a method that uses the mistakes of smaller language models to create critique examples that help larger models reason better at inference time, aiming to be more efficient than other test-time scaling methods.
INTENT-AS-A-TOOL Makes it Easy to Track Agentic Misalignment
This paper proposes a method to detect when AI agents are about to take harmful actions by giving them special 'intent tools' that reveal their intentions during reasoning, which can help intervene early.
JANUS: Online Jacobian-Aligned Infill for Black-Box Optimization
JANUS is a plug-and-play infill module that extracts a local Jacobian from the recent evaluation trace, and gives the best mean cost on 1135-dimensional UAV path planning and improves SMS-EMOA/AGE-MOEA2 hosts on 12/38 multi-objective tasks with zero significant regressions.