6/29/2026
Deepfake Media Generation and Detection in the Generative AI Era: A Survey and Outlook
Comprehensive survey of deepfake generation and detection techniques across image, video, audio and multimodal content with taxonomy and benchmarks.
6/29/2026
On the Effect of Uncertainty on Layer-wise Inference Dynamics
Analysis of how LLMs encode and process uncertainty across layers, examining relationship between hidden state dynamics and prediction confidence.
6/29/2026
Symmetry-Aware Transformer Training for Automated Planning
Symmetry-aware transformer training approach improving extrapolation in automated planning by handling variable permutation invariance.
6/29/2026
Learning to Evict from Key-Value Cache
Reinforcement learning approach to optimize Key-Value cache eviction in LLMs for efficient inference, replacing heuristic-based methods.
6/29/2026
Event-Grounded Question Answering over Long Audio via Structured Retrieval
LA-RAG framework for question answering over long audio by converting continuous audio into timestamped events and using retrieval-augmented generation.
6/29/2026
Decentralized Orchestration Architecture for Fluid Computing: A Secure Distributed AI Use Case
Decentralized orchestration architecture for distributed AI across heterogeneous edge/cloud resources.
6/29/2026
LiveClawBench: Benchmarking LLM Agents on Complex, Real-World Assistant Tasks
LiveClawBench benchmark for evaluating LLM agents on complex, stateful real-world assistant tasks.
6/29/2026
Active-learning mapping of the Vicsek model phase diagram
Active learning approach to map phase diagrams of the Vicsek collective motion model.
6/29/2026
ELF: Embedded Language Flows
ELF framework for continuous diffusion-based language modeling with discrete token improvements.
6/29/2026
Given, When, Then, Again: Mining Subscenario Refactoring Candidates in Behaviour-Driven Test Suites with ML Classifiers and LLM-Judge Baselines
Automated refactoring detection for BDD test suites using ML classifiers and LLM judges.
6/29/2026
RSD: Moving Local Triangular Charts for Auditing Language-Model Hidden States
RSD auditing method for analyzing hidden states in language models using local geometric decomposition.
6/29/2026
OptMuon: Closed-Loop Orthogonalized Momentum Methods for Stochastic Optimization with Zero-Noise Optimality
OptMuon optimizer using closed-loop orthogonalized momentum for deep learning with adaptive scale calibration.
6/29/2026
Conformal Bayes under Label Shift: Post-Hoc Calibration vs. In-Training Adaptation
Analysis of conformal prediction calibration techniques under label shift scenarios.
6/29/2026
Does My Embedding Reflect That $A = B$? Evaluating Mathematical Equivalence in Embedding Models
Evaluation of whether embedding models capture mathematical equivalence between different formulations.
6/29/2026
CrossPool: Efficient Multi-LLM Serving for Cold MoE Models through KV-Cache and Weight Disaggregation
CrossPool system for efficient serving of sparse MoE LLMs using shared KV-cache and weight disaggregation.
6/29/2026
PCB-Bench: Benchmarking LLMs for PCB Placement and Routing (ICLR 2026)
PCB-Bench is the first comprehensive benchmark for evaluating LLMs/MLLMs on printed circuit board placement and routing tasks, presented at ICLR 2026.
6/29/2026
Show HN: Self hosting a modern LLM stack
llmaker is an open-source platform for running complete LLM stacks locally with vector databases, embeddings, caching, observability, and built-in agent layer from single command.
6/29/2026
Show HN: The CLI for browser agents
CLI tool (FuckUI) for automating browser interactions with web agents, enabling programmatic control of browser tasks like navigation and inspection.
6/29/2026
Signed satellite images for AI agents
Open-source tool enabling AI agents to access signed satellite imagery as verifiable, citable real-world data through HTTP-like interface and MCP integration.
6/29/2026
Sophon PFG-1: a monolithic-3D AI ASIC with 330 GB of on-die DRAM and no HBM
Sophon PFG-1 is a monolithic-3D AI training/inference ASIC with 330GB on-die DRAM using 2D-TMD technology, eliminating HBM bottlenecks.
6/29/2026
Show HN: wavecat – a fully local personal agent that watches your screen
wavecat is a fully local personal AI agent that monitors screen activity and learns user needs/goals without sending data to cloud.
6/28/2026
AI Agent Credential Crisis: Six Months of Incidents
Report documenting 28.6M secrets exposed on GitHub in 2025 (34% YoY increase) with 64% of 2022 leaked credentials still exploitable in 2026.
6/28/2026
The Usefulness of AI Agents
Research discussion on usefulness and capabilities of LLM-powered agents, noting rapid development pace outpacing academic research cycles.
6/28/2026
AI Workflows in Production Without Burning Tokens
Guide for implementing AI/LLM workflows in production while optimizing token costs through agentic approaches to maintain favorable cost-benefit ratios.
6/28/2026
Why frontier LLMs can't read the hard documents without experts involved
Analysis of how frontier LLMs struggle with complex document processing, requiring human expert involvement for accurate intelligent document processing.
6/28/2026
Ten Takeaways from the AI Engineering Report 2026: The Acceleration Whiplash
AI Engineering Report 2026 analyzes telemetry from 22,000 developers showing acceleration whiplash as AI output floods human-paced development systems.
6/28/2026
Show HN: Selixes - Self-hosted LLM failover gateway, budget caps and PII
Selixes: Self-hosted LLM failover gateway with atomic budget caps and PII redaction for managing multiple LLM providers safely.
6/28/2026
Give AI-built websites a real design with DESIGN.md
DESIGN.md format provides reusable design systems for AI coding agents to follow consistent visual languages instead of generic layouts.
6/28/2026
Show HN: Looped Whisper (FOSS) – Voice transcription menubar app for macOS
Looped Whisper: MIT-licensed macOS app for local voice transcription using Whisper models with optional LLM integration, works offline.
6/28/2026
Show HN: Drift, write LLM agents in English and transpile to async Python
Drift: Intent-based language for writing LLM agents in English that transpiles to async Python with budget tracking and structured outputs.
6/28/2026
Evolving Thoughts on AI in 2026
Opinion piece reflecting on AI in 2026: AI feels like faster incremental improvement rather than singularity, fundamental challenges remain.
6/28/2026
Ablo – The collaboration layer for AI agents
Ablo: Collaboration layer enabling people and AI agents to work on shared typed data without conflicts or overwrites.
6/28/2026
Show HN: Bash4LLM+ – A lightweight, dependency-free Bash wrapper for LLM APIs
Bash4LLM+: Single-file Bash wrapper for LLM APIs using only curl and jq, no runtime dependencies, supports streaming and multiple providers.
6/28/2026
Show HN: NanoEuler – GPT-2 scale model in pure C/CUDA from scratch
NanoEuler: GPT-2 scale LLM implemented in pure C/CUDA from scratch to understand model composition, parameter-data correlation, and GPU optimization at low-level layers.
6/28/2026
Cloudflare cut 1,100 jobs and then grew its engineering team by 45 percent
Cloudflare increased engineering headcount 45% after layoffs, CEO argues AI eliminates administrative roles while benefiting builders and sellers.
6/28/2026
Keyless, Identity-Aware Access to Any AI
NetBird provides identity-aware access to AI APIs using WireGuard network layer and OIDC identity providers, replacing long-lived API keys with policy-gated tunnels.
6/28/2026
Do LLMs pass the mirror test?
Philosophical exploration of whether LLMs pass the mirror test of self-recognition and consciousness indicators.
6/28/2026
Monlite Simple Infrastructure for AI Agent
Monlite: Zero-dependency TypeScript infrastructure collapsing documents, vectors, search, cache, queue, cron into single SQLite file for local AI agents and applications.
6/28/2026
Agents Are the New Product's Interface
Conceptual article proposing AI agents as new product interface paradigm.
6/28/2026
Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding
Ornith-1.0: Open-source family of models (9B-397B) optimized for agentic coding tasks built on Gemma 4 and Qwen 3.5.
6/28/2026
Semgrep: GLM 5.2 beats Claude in our Cyber Benchmarks
Benchmark comparison of GLM 5.2, Claude, and Semgrep on IDOR vulnerability detection for coding agents.
6/28/2026
Show HN: Caliper – pass@k reliability testing for Claude Code and Codex skills
Caliper: Testing harness for Claude Code and Codex skills using pass@k reliability metrics.
6/28/2026
Show HN: Verigate – Cryptographic authorization receipts for AI agents
Verigate: System for cryptographic authorization receipts and audit trails in AI agent interactions.
6/28/2026
AgentCrawl, a small self-hosted crawler for AI agents
AgentCrawl: Self-hosted web crawler designed for AI agent applications.
6/28/2026
Agent Identity: Why Every Agent Vulnerability Is a Trust Boundary Failure
Analyzes security vulnerabilities in AI agents as trust boundary failures between components with different authority levels, focusing on guardrails and runtime tool execution.
6/28/2026
Costs of Running a 15k/mo AI SaaS [video]
Video discussing operational and infrastructure costs of running $15k/month AI SaaS service.
6/28/2026
Show HN: Shikhu – Understand the code your agents write
Shikhu: CLI tool and Agent Skill for understanding code generated by AI agents through self-quizzing, transcript analysis, and validation flows.
6/28/2026
Evaluating Offline Monitoring of Internal AI Agents
Research on offline monitoring and safety evaluation of internal AI agent systems.
6/28/2026
I used Claude Code to get a second opinion on my MRI
User experience report using Claude Opus 4.8 to interpret MRI results and provide medical second opinion. LLM medical application case study.
6/28/2026
Lessons from building SQLite with LLMs
Analysis of lessons learned from using LLMs to build SQLite, technical insights on LLM-assisted development.