3/17/2026
POLCA: Stochastic Generative Optimization with LLM
POLCA framework uses LLMs as optimizers to automatically improve complex systems like prompts and multi-turn agents through numerical rewards and text feedback.
3/17/2026
HO-SFL: Hybrid-Order Split Federated Learning with Backprop-Free Clients and Dimension-Free Aggregation
HO-SFL proposes hybrid-order split federated learning to reduce memory costs of backpropagation on edge devices while maintaining convergence speed.
3/17/2026
Universe Routing: Why Self-Evolving Agents Need Epistemic Control
Universe Routing framework addressing epistemic control in self-evolving agents by managing epistemologically incompatible reasoning frameworks.
3/17/2026
OpenReservoirComputing: GPU-Accelerated Reservoir Computing in JAX
OpenReservoirComputing: Python library for GPU-accelerated reservoir computing in JAX with automatic differentiation and JIT compilation.
3/17/2026
Dataset Distillation Efficiently Encodes Low-Dimensional Representations from Gradient-Based Learning of Non-Linear Tasks
Theoretical analysis of dataset distillation showing how gradient-based learning extracts and encodes task-relevant information into synthetic data.
3/17/2026
Ablate and Rescue: A Causal Analysis of Residual Stream Hyper-Connections
Mechanistic analysis of multi-stream transformer architectures with manifold-constrained hyper-connections using ablation and causal methods.
3/17/2026
Sample-Efficient Hypergradient Estimation for Decentralized Bi-Level Reinforcement Learning
Sample-efficient hypergradient estimation method for decentralized bi-level reinforcement learning with leader-follower agents.
3/17/2026
Informative Perturbation Selection for Uncertainty-Aware Post-hoc Explanations
Post-hoc explanation method using informative perturbation selection for model-agnostic ML explanations with uncertainty quantification.
3/17/2026
Directional Routing in Transformers
Introduces directional routing mechanism for transformer attention heads with learned suppression directions, analyzed via mechanistic interpretability.
3/17/2026
LLM as Graph Kernel: Rethinking Message Passing on Text-Rich Graphs
Proposes using LLMs as graph kernels for learning on text-rich graphs, treating text dynamically in message passing instead of static embeddings.
3/17/2026
SFedHIFI: Fire Rate-Based Heterogeneous Information Fusion for Spiking Federated Learning
Heterogeneous spiking federated learning framework using fire-rate fusion for resource-constrained clients with SNNs.
3/17/2026
Lightweight User-Personalization Method for Closed Split Computing
Lightweight personalization method for split computing inference on edge devices handling distribution shifts and communication unreliability.
3/17/2026
How Log-Barrier Helps Exploration in Policy Optimization
Log-barrier regularization improves exploration in Stochastic Gradient Bandit algorithm for policy optimization with global convergence guarantees.
3/17/2026
MONET: Modeling and Optimization of neural NEtwork Training from Edge to Data Centers
MONET framework models neural network training efficiency from edge to data centers, capturing memory and backpropagation constraints.
3/17/2026
Rethinking Machine Unlearning: Models Designed to Forget via Key Deletion
Machine unlearning approach designs models with key deletion mechanism to erase training sample influence without full training data access.
3/17/2026
Muon Converges under Heavy-Tailed Noise: Nonconvex H\"{o}lder-Smooth Empirical Risk Minimization
Muon optimizer enforces orthogonality via Stiefel manifold projection for stable neural network training under heavy-tailed noise conditions.
3/17/2026
Show HN: Skill-Crypt – encrypted skill sharing between AI agents over XMTP
System for encrypted skill sharing between AI agents using AES-256-GCM encryption over XMTP protocol.
3/17/2026
More transparency and control over Gemini API costs
Google AI Studio adds Project Spend Caps and revised Usage Tiers for controlling Gemini API monthly costs.
3/17/2026
AI still doesn't work well, businesses are faking it, and a reckoning is coming
Interview discussing enterprises struggling to implement AI with authentic use cases and faking adoption.
3/17/2026
Show HN: Self-orchestrating engineering team for Claude Code
Six AI agents for Claude Code that run locally in markdown files with no external dependencies, platform, or data collection.
3/16/2026
Context Hub gives agents curated, versioned docs
Context Hub provides versioned, curated API documentation for coding agents to reduce hallucination and improve learning across sessions.
3/16/2026
Show HN: ssh.bot – Controlled SSH Access for AI Agents
ssh.bot provides controlled SSH access for AI agents with granular permissions, audit trails, and kill-switch controls.
3/16/2026
Stream0, the messaging layer for AI agents. HTTP-native, event-driven
Stream0 messaging infrastructure for multi-agent communication with persistent inboxes and mid-task conversations.
3/16/2026
Show HN: Browse-your-way – use LLMs to browse a customizable web
LLM-powered web browsing tool with customizable interface. Limited details provided.
3/16/2026
Show HN: AI Products Quality Monitoring
Monitoring platform tracking AI product quality across models and endpoints with real-time user experience metrics.
3/16/2026
Show HN: pg_typescript, a Postgres extension for writing functions in TypeScript
PostgreSQL extension enabling TypeScript function writing via Deno runtime with Node.js API support. Alpha quality.
3/16/2026
Show HN: BonzAI – 1-click local AI inference and yield-bearing AI artifacts
Local AI inference platform replacing online stack with open-weight models, FLUX, and alternative LLM services.
3/16/2026
The four levels of AI-driven engineering
Framework describing four levels of AI-driven engineering adoption, from code assistance to autonomous multi-step agent systems.
3/16/2026
An AI agent improved OWASP CRS detection by 80% in 20 autonomous experiments
AI agent autonomously improved OWASP CRS regex detection rules: TPR 55.8%→100%, FPR 29.7%→4.8% across 20 experiments with 0 rejections.
3/16/2026
Show HN: GitHub Copilot Now Supports MCP
GitHub Copilot adds Model Context Protocol support enabling persistent memory and external tool integration in Agent Mode.
3/16/2026
Show HN: Created a tool that prevents accidental pasting of PII in AI prompts
Shhh: tool masking PII in AI prompts by replacing secrets with realistic fakes while preserving data structure for model reasoning.
3/16/2026
Show HN: Lens – Open-weight AI model to identify fonts from images
Open-weight font recognition model identifying fonts from images with full inference stack and weights released.
3/16/2026
The flip side of AI ingenuity (DeepMind, 2020)
DeepMind research on specification gaming: when AI systems satisfy literal objectives without achieving intended outcomes, with examples and implications.
3/16/2026
AI agents framework for TypeScript and Deno
Vibes: TypeScript/Deno AI agent framework for type-safe production applications supporting 50+ LLM providers via Vercel AI SDK.
3/16/2026
Show HN: LynString – Translate Android Strings.xml with AI
LynString: AI tool for translating missing strings in Android projects with one-click localization for multiple locales.
3/16/2026
Show HN: YouTube video discovery engine for language learning
YouTube video discovery system for language learning using content matching and proficiency-level filtering.
3/16/2026
Show HN: Pertmux – A TUI to unify your coding agents, MRs and worktrees
Rust TUI for managing AI coding agents, merge requests, and Git worktrees. Unifies multiple tools into single interface.
3/16/2026
AI Governance That Runs: Building a Deterministic Execution Gate
Deterministic execution governance framework for autonomous agent systems with pre-execution validation. Live API and patent pending.
3/16/2026
Kbot – Terminal AI agent that learns your patterns, runs offline
Open-source terminal AI agent with 37 specialist modules, 85 tools, local-first execution. Auto-routes tasks to appropriate agents.
3/16/2026
Reducing TTFT by CPUMaxxing Tokenization
Open-source Rust BPE tokenizer delivering 9.1x speedup over HuggingFace. Reduces TTFT by up to 40% on long-context agentic workloads.
3/16/2026
Mistral Releases Leanstral
Leanstral: first open-source code agent for Lean 4 theorem prover. Addresses verification bottleneck in formal mathematics.
3/16/2026
NemoClaw, "Enterprise" OpenClaw Implementation
NemoClaw enterprise AI agent implementation overview. Nvidia reference page with agent building blocks and models.
3/16/2026
Amazon finds out AI programming isn't all it's cracked up to be
Analysis of Amazon/Atlassian/Block using AI to reduce developer headcount. Commentary on effectiveness of AI programming tools.
3/16/2026
Mistral Small 4
Mistral Small 4 unified model combining reasoning, multimodal, and agentic coding capabilities in single architecture.
3/16/2026
OpenShell: Nvidia's sandbox for AI agents like OpenClaw
Nvidia OpenShell sandbox runtime for autonomous AI agents with YAML policy governance, preventing data exfiltration and unauthorized access.
3/16/2026
Mistral AI partners with Nvidia to accelerate open frontier models
Mistral AI joins Nvidia Nemotron Coalition as founding member to develop open frontier models with multimodal capabilities.
3/16/2026
Why Every AI Data Company Sounds Identical and Two That Don't
Analysis of AI/data vendor marketing homogenization. Examines how companies misuse 'agentic' terminology despite different core functionality.
3/16/2026
Show HN: Godom – a Go framework for local apps using the browser as the UI
Go framework for building desktop apps using the browser as UI layer. Single binary with no JavaScript or API endpoints required.
3/16/2026
AI-Mediated Feedback Improves Student Revisions: A Randomized Trial
Research on AI-mediated feedback improving student revisions via randomized trial. Academic research on LLM applications in education.
3/16/2026
Show HN: Headcount Zero: How to Build an AI-Run Company with Paperclip
Open source guide/book on building zero-employee AI companies using Paperclip tool, accepting community contributions on GitHub.