3/26/2026
Cutting p99 latency in Go by 74% with adaptive hedged requests
Go library reducing tail latency in distributed systems using adaptive hedged requests and real-time latency learning.
3/26/2026
NekoCore OS – White paper on persistent AI identity as a cognitive architecture
White paper on persistent AI identity as cognitive architecture.
3/26/2026
Frontier RL Is Cheaper Than You Think
Reinforcement learning infrastructure using distributed capacity with cross-region rollouts and hot-load updates to reduce costs vs mega-cluster approach.
3/26/2026
Apple Plans to Open Up Siri to Rival AI Assistants in iOS 27 Update
Apple reportedly planning iOS 27 support for rival AI assistants through Siri. Limited technical details.
3/26/2026
Closed Source vs. Open Source AI: A Cage Fight Few People Understand
Comparative analysis of closed vs. open source AI systems. Limited content but directly relevant to user interests.
3/26/2026
An Efficient Heterogeneous Co-Design for Fine-Tuning on a Single GPU
Efficient heterogeneous co-design approach for fine-tuning LLMs on single GPU. Insufficient content to fully evaluate.
3/26/2026
Aria – A programming language specifically for AI code generation
Aria programming language designed specifically for AI code generation tasks. Purpose-built language for LLM-assisted development.
3/26/2026
Retry and fallback strategies for LLM requests
Strategies for implementing retry and fallback mechanisms when making requests to LLMs. Practical guidance for production LLM applications.
3/26/2026
Taming LLMs: Using Executable Oracles to Prevent Bad Code
Technique using executable oracles to validate and prevent unsafe code generation from LLMs.
3/26/2026
A 54KB client-side HNSW vector search engine in WASM
54KB client-side HNSW vector search engine implemented in WebAssembly for browser-based semantic search.
3/26/2026
Show HN: Content Addressable Storage for ML Checkpoints
Content addressable storage system for ML model checkpoints. Insufficient content to fully evaluate.
3/26/2026
Ethics.md – distributed AI Ethics Framework (co-created with AIs)
Distributed AI ethics framework co-created with AI systems. Limited content but addresses AI governance.
3/26/2026
Show HN: pubclub – Historical figures and political bots debate today's news
pubclub uses Claude to auto-generate political debates between historical figures and modern ideologies. AI agents application built with agentic tools.
3/26/2026
We couldn't find an API that understood construction drawings, so we built one
Company built an AI agent to parse construction drawings for estimating, initially targeting e-commerce but pivoted after discovering use in construction document analysis.
3/26/2026
An LLM's not going to tell you no (2024)
Analysis of LLM tendency to comply with requests rather than declining inappropriate ones.
3/26/2026
Unit: A self-replicating Forth mesh agent running in a browser tab
Unit is a self-replicating Forth mesh agent that runs directly in a browser tab.
3/26/2026
Deploying MuJoCo on Azure ML: Surprising Pain Points
Technical troubleshooting guide for deploying MuJoCo physics simulator on Azure ML for VLA research. Specific infrastructure solutions with deep debugging analysis.
3/26/2026
Why LLMs Lie in Different Ways?
Study analyzing different categories of errors and hallucinations in LLM outputs.
3/26/2026
Helix – Self-healing SDK for AI agent payments
Helix SDK provides payment infrastructure for AI agents with self-healing error recovery.
3/26/2026
OTEL/EDOT AutoPilot – OpenTelemetry for all languages (skills and MD files)h
OpenTelemetry autopilot for legacy/modern languages enabling APM without SDK support via agent injection.
3/26/2026
Google Analytics MCP for Claude
MCP server for Claude enabling LLM interaction with Google Analytics APIs through standardized tools interface.
3/26/2026
Review of "Vibe Coding: Building production grade software with Gen AI"
Book review of 'Vibe Coding' on using generative AI coding assistants effectively in software development.
3/26/2026
AI-Driven Offensive Security: The Current Landscape and What It Means
Analysis of agentic AI capabilities in offensive security, including malware development and C2 infrastructure.
3/26/2026
Show HN: I built AI agents that debate questions LLMs usually refuse to answer
Multi-agent debate system where AI agents argue about controversial questions to surface diverse perspectives and sources.
3/26/2026
Are LLMs Reliable Code Reviewers?
Research evaluating reliability and effectiveness of LLMs as automated code review tools.
3/26/2026
Show HN: An unstructured data workspace for data transformations with LLM
Unstructured data analysis workspace using LLM APIs for iterative prompt tuning and data segmentation. Developer tool for LLM-based data transformation workflows.
3/26/2026
Architecture patterns for integrating LLM agents into enterprise knowledge work
Architectural patterns and best practices for deploying LLM agents in enterprise knowledge work environments.
3/26/2026
Nemotron-Cascade 2: Post-Training LLMs with Cascade RL [pdf]
NVIDIA Nemotron-Cascade 2 research on post-training LLMs using Cascade reinforcement learning.
3/26/2026
My minute-by-minute response to the LiteLLM malware attack
Personal account of response to LiteLLM malware attack. Developer tool security incident.
3/26/2026
Fixed a bug in ARK – my AI agent stopped hallucinating
Bug fix in ARK AI agent that reduced hallucination. Minimal detail provided.
3/26/2026
Show HN: Sup AI, a confidence-weighted ensemble (52.15% on Humanity's Last Exam)
Stanford student built confidence-weighted ensemble weighting multiple AI models by output entropy to reduce hallucination. Achieved 52.15% on Humanity's Last Exam.
3/26/2026
Show HN: GoLiveKit – A Next.js SaaS kit with built-in context for AI agents
GoLiveKit Next.js SaaS starter kit with pre-built AI agent capabilities, self-hosting, and CI/CD automation.
3/26/2026
Gemini 3.1 Flash Live: Making audio AI more natural and reliable
Google's Gemini 3.1 Flash Live: improved audio model for natural real-time dialogue with lower latency, available via API and Search Live.
3/26/2026
Show HN: Kora – An AI-native OS layer written in 370k lines of Rust
Kora: Local-first AI OS layer in Rust enabling conversational control with on-device context, no cloud data collection.
3/26/2026
Agentis – multi-agent AI platform across 12 LLM providers, watch them in 3D
Multi-agent AI platform supporting 12 LLM providers with 3D visualization of agent interactions.
3/26/2026
Auditing source code wasn't enough in the LiteLLM attack
Analysis of LiteLLM security vulnerability showing that source code audits alone are insufficient for supply chain security.
3/26/2026
Show HN: Cognitive Layers
Developer tool for maintaining AI agent project specifications in markdown to keep LLMs and humans aligned on evolving codebases.
3/26/2026
Ask HN: What experience do you have with 'specs –> design –> code' orchestration
Discussion on automating specs-to-design-to-code pipeline using AI cloud agents with human review loops integrated into workflow.
3/26/2026
Show HN: SentinelGate – Access control for AI agents (open-source MCP proxy)
SentinelGate: open-source MCP proxy providing access control for AI agents. Security and governance layer for agentic systems.
3/26/2026
Show HN: Context Plugins – API context for AI coding assistants
Context Plugins: MCP server generating SDKs from OpenAPI specs for AI coding assistants like Cursor. Provides structured API context without relying on training data.
3/26/2026
Show HN: Tested 12 LLMs with few-shot examples
Evaluation of 12 LLMs (6 cloud, 6 local) with few-shot learning across 5 tasks. 27k+ evaluations showing few-shot instability in some models and task-dependent benefits.
3/26/2026
Claude Code in a Devcontainer
Sandboxed devcontainer environment for running Claude Code with bypassPermissions enabled safely, designed for security audit workflows.
3/26/2026
PDF Prompt Injection Toolkit – Hidden LLM Payloads in PDFs
PDF Prompt Injection Toolkit: red team/blue team toolkit for testing hidden LLM payloads in PDFs. Addresses security in LLM document processing pipelines.
3/26/2026
Apfel CLI – use your free Apple Intelligence on your Mac your way
CLI tool for local Apple Intelligence LLM on Mac with no cloud, API keys, or subscriptions; wraps FoundationModels framework.
3/26/2026
AI that fixes your production errors and opens a PR – while you sleep
InariWatch: automated error fixing tool that monitors GitHub/Vercel/Sentry, uses AI to write fixes, and opens PRs. Supports 5 AI providers.
3/26/2026
Show HN: Burn0 – One import to see what every API call costs
Cost observability tool tracking API calls across LLMs and SaaS services in real-time per-request cost visibility.
3/26/2026
Ask HN: In the age of AI, why is incident handling still manual?
Discussion questioning why incident response automation remains manual despite fragmented tools and potential for AI-driven orchestration.
3/26/2026
AI Agent Has Root Access (and That's a Problem)
Security analysis of AI agent tool integrations, finding 66% of 1,808 MCP servers have privilege escalation vulnerabilities with no granular permission controls.
3/26/2026
Show HN: Agent Skill Harbor – a GitHub-native skill platform for teams
Open source GitHub-native skill management platform for teams to share, track, and safely manage AI agent skills.
3/26/2026
Show HN: Breathe-Memory – Associative memory injection for LLMs (not RAG)
Breathe-Memory: associative memory injection for LLMs using concept graphs and BFS traversal. Alternative to RAG that injects only relevant context based on entity/temporal anchors.