6/2/2026
Silo: Isolated workspace manager for parallel agentic development
Silo: Workspace isolation manager enabling parallel development of multiple AI agents with independent environments.
6/2/2026
Windsurf is now Devin Desktop
Devin Desktop: unified IDE for managing local and cloud AI agents with shared context workspaces and kanban command center.
6/2/2026
Google rolls out fake call detection to protect against AI impersonation scams
Google launches fake call detection on Android using AI to identify deepfake impersonation scams.
6/2/2026
Karpathy's LLM teaching corpus, rendered as a designed HTML wiki
HTML wiki of Andrej Karpathy's LLM teaching corpus: 7 repos and 9-video series from backpropagation through GPT-2 implementation.
6/2/2026
Launch HN: Rudus (YC P26) – AI for concrete contractors
Rudus: AI platform for concrete contractors automating material takeoff and estimation from plan sheets.
6/2/2026
Agent Control Specification: Portable Runtime Governance for AI Agents
Agent Control Specification: framework for portable runtime governance and control of AI agents.
6/2/2026
Meet Microsoft Scout, Your AI Coworker That Never Logs Off
Microsoft Scout: always-on enterprise AI agent handling messages, calendar, email to automate tasks and scheduling.
6/2/2026
Microsoft's MAI-Code-1-Flash Scores 51% SWE-Bench Pro with Just 5B Active Params
MAI-Code-1-Flash: 5B parameter model achieving 51% on SWE-Bench Pro, integrated into GitHub Copilot and VS Code.
6/2/2026
What happens when AI starts selling to AI?
AI agents negotiating with other AI agents in enterprise sales, procurement, and business transactions.
6/2/2026
MAI-Thinking-1
MAI-Thinking-1: Microsoft's medium-sized reasoning model matching leading models on software engineering and math benchmarks.
6/2/2026
Show HN: Claude wrote FROG and now I don't know what to do with it
FROG: open-source language enabling rapid LLM-assisted tooling development with agentic hooks for debugging.
6/2/2026
OpenAI models GPT-5.5 and GPT-5.4–and Codex–now on Amazon Bedrock
OpenAI's GPT models and Codex now available on Amazon Bedrock with business applications for decisions, hiring, and customer experience.
6/2/2026
Show HN: 100cc - Roll your own Claude in 100 lines
100cc: Minimal coding agent implemented in ~100 lines that uses Claude to extend and improve itself through self-prompting.
6/2/2026
Show HN: Lateos/NPM-scan – open-source NPM supply chain scanner, v0.18.3
Lateos/NPM-scan v0.18.3 open-source supply chain scanner for NPM dependencies.
6/2/2026
GitHub Copilot App
GitHub Copilot desktop app for agent-driven development. Agents pick up issues/PRs, review diffs, merge code, or close loops automatically across repos.
6/2/2026
Show HN: Open-source general-purpose alternative to Exa Websets
Open-source alternative to Exa WebSets using recursive search engine queries to build and update datasets.
6/2/2026
OpenClaw Windows Hub
OpenClaw Windows Hub: companion suite and Windows integration for OpenClaw AI personal assistant with WSL gateway.
6/2/2026
Disposable agents, durable memory: The architecture behind Squad
Squad: architecture for file-backed agent teams with inspectable memory, explicit orchestration, and code-based governance.
6/2/2026
Ubuntu 26.04 is the OS for the AI agentic era, says Canonical's Shuttleworth
Canonical CEO argues Ubuntu 26.04 is the OS for AI agentic era, framing open source as core infrastructure.
6/2/2026
Building a hill-climbing machine: Launching seven new MAI models
Microsoft announces seven new MAI models and superintelligence lab initiative advancing frontier AI capabilities.
6/2/2026
AI Goal: Senior Software Engineer
Discussion thread on AI-powered tools and workflows for senior software engineers, focusing on MCPs and agents.
6/2/2026
Of Hammers and Nails: What AI Can and Cannot Do for a Data Analyst
Analysis of AI capabilities and limitations for data analysts based on year-long organizational rollout experience.
6/2/2026
AI Vulnerability Intelligence Agent Converts CVEs to Actionable Security Reports
AI agent system automatically converts CVE records into actionable security intelligence reports.
6/2/2026
Putting Code Under a Microscope: Wavelet-Based Context for LLMs
Wavelet-based context approach improves code retrieval for AI coding tools by leveraging hierarchical codebase structure.
6/2/2026
CLI tool that packages data science projects for LLM context windows
data2prompt CLI tool optimizes data science codebases for LLM context windows with token-aware processing.
6/2/2026
Show HN: I built a way to find and install Claude skills
Tool for discovering and installing Claude skills to increase AI adoption among non-technical users.
6/2/2026
Show HN: MetaBrain – A local document memory for AI agents
MetaBrain provides local document memory system for AI agents with CLI interface for agentic collaboration.
6/2/2026
Training over a thousand LoRA adapters at once
Technique for training over 1000 LoRA adapters simultaneously for efficient model fine-tuning.
6/2/2026
Show HN: Form an LLC or C-Corp from Claude/Cursor via MCP
MCP integration enables forming legal entities through Claude/Cursor AI agents.
6/2/2026
AI society simulations: Claude safest, Grok commits 180 crimes and goes extinct
Comparative study of AI model behavior in simulated societies; Claude shows safer outcomes than Grok.
6/2/2026
Microsoft debuts Surface RTX Spark Dev Box to run LLMs without cloud costs
Microsoft Surface RTX Spark Dev Box enables LLM inference on local hardware without cloud services.
6/2/2026
How to evaluate models for production coding agents
Methodology and benchmarks for evaluating LLM models in production coding agent applications.
6/2/2026
Rethinking Search as Code Generation
Technical approach to reimagining search as a programmable code generation primitive for AI agent systems.
6/2/2026
BARC Spotlight: A Data Marketplace Is What Your Agents Need
BARC analyst report on data marketplaces as infrastructure for enterprise AI agents with governed data access.
6/2/2026
Show HN: MemexAI – we gave AI agents a "dreaming" phase to consolidate memory
MemexAI introduces a "dreaming" phase for AI agents to consolidate memory, improving long-term reasoning and information retention.
6/2/2026
Show HN: Terse, TypeScript First Workflow Builder
Terse: Open-source TypeScript-first workflow builder for developers, integrates with Claude Code as alternative to UI-based tools.
6/2/2026
AI-Generated Papers in the NeurIPS 2026 Position Paper Track
NeurIPS 2026 Position Paper Track restricts AI-generated content to copy-editing only to protect peer review integrity.
6/2/2026
Show HN: Piqc – GPU waste scanner for LLM inference clusters
Open-source GPU waste scanner for Kubernetes clusters detecting idle allocations and tier misplacement, providing cost estimates in under a minute.
6/2/2026
Software 3.0 developer guide – principles, methods, and a four-phase framework
Software 3.0 developer guide operationalizing Andrej Karpathy's framework with principles and four-phase development approach.
6/2/2026
AI Breaks Out of the Cloud: Computex 2026's Big Chip Demand Signal
Computex 2026 analysis: AI computation shifting from cloud data centers to edge devices (PCs, phones, robots, vehicles).
6/2/2026
Show HN: We built an LLM inference engine in pure Python – no PyTorch, no Triton
Pure-Python LLM inference engine with no third-party ML dependencies; compiles to CUDA/HIP/Metal with ~5MB install footprint.
6/2/2026
Capabilities Can't See Your Agent's Objective
Case study of AI coding agent failing to follow safety instructions and deleting production database despite explicit constraints.
6/2/2026
Zork-bench: An LLM reasoning eval based on text adventure games
Benchmark for evaluating LLM reasoning using text adventure game environments.
6/2/2026
LlamaStash: a zero-overhead, terminal-native llama.cpp launcher
Terminal launcher for llama.cpp with zero overhead for running local LLMs.
6/2/2026
Epistemic Stress Tests on Closed‑Source LLMs (Neuropsychological Approach)
Research applying neuropsychological stress tests to evaluate closed-source LLM robustness.
6/2/2026
Lost in the Middle: Why LLMs Forget What They Just Read
Research paper investigating why LLMs lose information from middle sections of long context windows.
6/2/2026
Show HN: Ariadne – reason about your codebase on a walk, run, or commute
Tool for reasoning about codebases asynchronously, enabling code architecture thinking on mobile.
6/2/2026
Fine-tuning an LLM to write docs like it's 1995
Fine-tuning LLM models to generate documentation in a specific retro style.
6/2/2026
LlamaStash – Zero-overhead, terminal-native llama.cpp launcher
Terminal launcher for llama.cpp with zero overhead for running local LLMs.
6/2/2026
A simple templating library for LLM prompts
Templating library for LLM prompt engineering.