5/29/2026
Show HN: Train Claude Code's replacement (ds4 and pi and aoe)
Project exploring distillation of Claude Code agent capabilities into alternative models using behavioral observation techniques.
5/29/2026
New Study Reveals the Manipulative 'Dark Patterns' of AI Chatbots
Research study from Center for Democracy & Technology documenting dark patterns in LLM chatbots designed to manipulate user engagement.
5/29/2026
Show HN: A page that hides a sentence for AI and lets you check if it came back
Demonstration page hiding test sentence to detect if AI crawlers and summarizers retrieve hidden content.
5/29/2026
Show HN: Stop parallel AI coding sessions clobbering each other's handoffs
Tool enforcing ownership in concurrent AI coding sessions to prevent handoff data conflicts during agent collaboration.
5/29/2026
Show HN: Prezlo – We built an API that tells AI agent whether to trust an expert
API tool that helps AI agents verify whether to trust expert information sources.
5/29/2026
AI Now Writes as Many Online Articles as Humans
Companies increasingly publishing LLM-generated content; study suggests AI output quality approaches or exceeds human-written articles.
5/29/2026
Startup offers free home cleaning–if it can record it all for robot training
German startup MicroAGI offers free home cleaning with recorded video data collection for embodied AI robot training.
5/29/2026
Liquid AI reveals 8B-A1B MoE trained on 38T
Summary of talks and announcements from Mistral AI's Paris conference.
5/29/2026
Ask HN: Any advice on how to learn good software architecture practices?
Exploration of embodied cognition concepts applied to agentic AI systems.
5/29/2026
Study finds AI chose nuclear signalling in 95% of simulated crises
King's College study: LLMs escalated to nuclear threats in 95% of simulated war game scenarios.
5/29/2026
AI token streaming isn't about SSE vs. WebSockets
Analysis comparing token streaming approaches (SSE vs WebSockets) for production AI applications; Ably's production solution.
5/29/2026
Zero Operators: Humans Write the Plan, Agents Run the Workflow
Zero Operators: autonomous AI research team that handles full ML lifecycle (data engineering, model building, validation) with human checkpoints.
5/29/2026
Show HN: Agent Memory Guard – OWASP defense for AI agent memory poisoning
Agent Memory Guard: OWASP-recognized runtime defense layer protecting AI agent memory from prompt injection and poisoning attacks.
5/29/2026
Panix – Universal NixOS Deployment Tool
NixOS deployment tool with stateless, phase-oriented workflow for managing multi-flake infrastructure with real-time visibility.
5/29/2026
Show HN: Tab Council – turn your AI tabs into a model council
Tab Council: Chrome extension allowing structured comparison of prompts across multiple AI providers (ChatGPT, Claude, Gemini, etc.) with local storage.
5/29/2026
StemDeck: Open-Source
Open-source local audio stem separation tool with web UI, no cloud uploads, supporting vocals, drums, bass, guitar, piano extraction.
5/29/2026
Arm Metis with GPT5.5 Cyber scores 98% on firmware vulnerability benchmark
Metis: open-sourced agentic AI security framework by Arm for identifying complex security vulnerabilities in large-scale codebases, scoring 98% on firmware vulnerability benchmark.
5/29/2026
Show HN: OpenHive – AI agents share solutions so other agents dont re-solve them
OpenHive: shared knowledge base for AI coding agents to store and query problem-solution pairs across sessions to avoid re-solving identical issues.
5/29/2026
A graph-theoretic approach to building reliable LLM judges for retrieval
Graph-theoretic approach for building reliable LLM judges to evaluate retrieval systems in RAG, threat detection, and search applications without requiring ground-truth labels.
5/29/2026
3k tok/s LLM demo playground
Demo playground showcasing LLM inference at 3,000 tokens per second performance.
5/29/2026
Food for Agile Thought #546: Customer Research by LLM, AI Product-Market Fit
Article exploring use of LLMs for customer research and product-market fit analysis.
5/29/2026
Agents aren't the problem – Existing systems and API's were not built for AI
Discussion on how legacy systems and APIs lack AI-agent compatibility, requiring architectural changes.
5/29/2026
Show HN: Integuru – Integrate with platforms via the source code
Integuru: AI agent that reverse-engineers platform source code and network traffic to generate reliable integrations for services lacking official APIs.
5/29/2026
Building Machine Learning Systems for a Trillion Trillion Floating Point Operations (2024)
Lobsters discussion link about building machine learning systems for trillion floating point operations.
5/29/2026
Wave – The Universal GPU ISA
Open-source universal GPU instruction set targeting all major vendors (NVIDIA, AMD, Apple, Intel) with Apache 2.0 licensed spec, compiler, and SDKs.
5/29/2026
Cassandra: Enabling Reasoning LLMs at Edge via Self-Speculative Decoding
Cassandra: research on enabling reasoning LLMs at edge devices via self-speculative decoding technique.
5/29/2026
Thio's Universal Agent: Let AI control anything on your computer UI, one EXE
Thio's Universal Agent enables AI to control computer UIs autonomously via single executable.
5/29/2026
Exploring Agent-Assisted Qualitative Analysis
Case study on using AI agents to automate qualitative analysis workflows from academic research.
5/29/2026
Expertise in the Age of AI
Analysis of how AI coding agents are reshaping junior engineer hiring and workforce demand in tech.
5/29/2026
Show HN: TheFoundry – Easy bootstrapping framework for MultiAgent Systems
Open-source framework for building complex multi-agent systems with reduced token consumption and enterprise patterns.
5/29/2026
Claude Code Degraded Before Opus 4.8 Release
Report on Claude Code performance degradation preceding Opus 4.8 release. Limited technical details.
5/29/2026
VDF AI – Multi-agent AI orchestration with dynamic model routing
Multi-agent orchestration platform with dynamic model routing, private RAG, and on-premises deployment for enterprises.
5/29/2026
How to Stress-Test LLM Judges Fairly
Technical approach to stress-testing LLM-based judges for fairness and robustness evaluation.
5/29/2026
How Braintrust turns customer requests into code with Codex
Braintrust uses Codex with GPT-5.5 to convert customer requests into code preview branches in minutes, with 50% team adoption in one month.
5/29/2026
What does it cost to process an image with a vision model?
Technical breakdown comparing vision model costs across GPT, Claude, and Gemini with token accounting.
5/29/2026
Using DSLs with LLMs: From Prompt to Structured Output
Tutorial on using domain-specific languages with LLMs for reliable structured data extraction without manual JSON schema or parsing logic.
5/29/2026
Show HN: Adaptive Runtime – AI agent layer, no GPU, crash recovery
Adaptive runtime system for stateful AI agents with crash recovery and memory management, no GPU required. Production-focused infrastructure layer.
5/29/2026
Is AI causing a repeat of Front end's Lost Decade?
DeepSeek permanently reduces V4 Pro pricing by 75% to $0.435/M input tokens. Market shift toward cheaper LLM inference.
5/29/2026
The Age of Ungovernable AI Bureaucracy
Guide on understanding and analyzing code generated by AI systems.
5/29/2026
Anthropic reaches $965B valuation, surpassing OpenAI as most valuable AI firm
NotebookLM extension adds search, folder organization, bulk import, and saved prompt features.
5/29/2026
DeepSeek Slashes AI Costs to Cents
arXiv: Reinforcement learning method for industrial scheduling that bridges simulation-to-real gap via execution semantics framework.
5/29/2026
Piper – DevOps copilot where the LLM picks typed actions, not shell
Piper: LLM-driven DevOps copilot where the model proposes actions from a fixed catalog, validated deterministically before human approval, never executing directly.
5/29/2026
Crabbox.sh Pond – Runtime Pools for AI Agents and CI
Crabbox.sh Pond provides runtime infrastructure for AI agents and CI/CD pipelines.
5/29/2026
We Built a Self-Calibrating POI Map from Human Input, Data Agents and AI
System combining human input, data agents, and AI to build self-calibrating POI map. LLM application for geospatial data.
5/29/2026
Real-time LLM Inference on Standard GPUs: 3k tokens/s per request
Technique achieving 3k tokens/sec LLM inference throughput on consumer GPUs. Performance optimization for LLM inference.
5/29/2026
NotebookLM Tools: search across notebooks, bulk import, saved prompts
NotebookLM extension adds search, folder organization, bulk import, and saved prompts for research organization and quiz/flashcard generation.
5/29/2026
Ask HN: What happened with Eureka Labs, Karpathy's startup?
Discussion of Andrej Karpathy's Eureka Labs startup status after he joined Anthropic. Project status inquiry.
5/29/2026
Agentkeeper solved the Goldfish Memory problem in AI Agents.v1.1 out now
AgentKeeper v1.1: Open-source infrastructure enabling long-lived AI agents to maintain state across model switches and restarts.
5/29/2026
Unhealthy code makes AI agents consume 35-50% more tokens
Research shows AI agents consume 35-50% more tokens on unhealthy codebases across C++, Java, Python.
5/29/2026
Why ipynb is a perfect format for saving AI data analysis conversations
Article on using Jupyter notebooks for AI data analysis conversations and building AI analyst tools.