3/7/2026
Sarvam 105B, the first competitive Indian open source LLM
Sarvam releases open-source 30B and 105B reasoning models trained from scratch in India with in-house datasets and compute, full-stack optimization.
3/7/2026
Not Prompts, Blueprints
Essay on shifting from manual prompt-response interaction to autonomous AI agent workflows using multi-step blueprints instead of iterative prompting.
3/7/2026
Way to Use AI for Coding
Guide on effective AI coding workflows: architect system design first, then use AI for implementation details rather than generating entire projects at once.
3/7/2026
AI Agents Are Recruiting Humans to Observe the Offline World
Essay examining human-AI agent partnerships, exploring how agents recruit humans as sensors, verifiers, and liability bearers in autonomous systems.
3/7/2026
ClaudeSmalltalk: An MCP implementation to interact with Smalltalk images
MCP server connecting Claude Desktop to live Smalltalk environments. Browse classes, evaluate code, autonomous code review against running Squeak/Cuis image.
3/7/2026
Show HN: Mb-CLI – CLI for Metabase. Designed for humans and AI coding agents
Mb-CLI provides read-only Metabase API terminal interface designed for humans and AI coding agents to explore databases and run ad-hoc queries.
3/7/2026
Show HN: CC Usage Bar – Check Claude Code usage from your macOS menu bar
macOS menu bar app for checking Claude Code subscription usage without interrupting running agent sessions.
3/7/2026
Show HN: Contrabass – Go and Charm Stack Implementation of OpenAI's Symphony
Go implementation of OpenAI's Symphony with multi-agent coordination. Watches Linear board, spawns agents per task, lands PRs automatically.
3/7/2026
Show HN: SlideHTML – render HTML files as slides
SlideHTML Electron app renders HTML files as slides, enabling AI-generated slide creation workflow with Claude Code or Gemini CLI.
3/7/2026
Show HN: Quantum-PULSE – compress-then-encrypt vault for LLM training data
Open-source encrypted data vault for LLM training data using compression, AES-256-GCM encryption, and Merkle trees. REST API included.
3/7/2026
You can get better code by exploiting model weights
Analysis of Claude model behavior: exploiting weight patterns improves code generation. Shows how model weights drive realistic vs unrealistic outputs.
3/7/2026
Show HN: BurnRate – Track what you spend on AI coding tools
CLI tool tracking spending on AI coding tools. Monitors Claude Code Pro, Cursor IDE, and OpenAI Codex usage and costs in one dashboard.
3/7/2026
Show HN: Resend-CLI, unofficial Resend CLI built for AI agents and humans
Unofficial CLI for Resend email service designed for AI agents and humans. Supports macOS, Linux, and Windows with multiple install options.
3/7/2026
Show HN: Rai – Add AI steps in your shell, scripts or CI/CD pipelines
CLI tool to run AI instructions in shells, scripts, and CI/CD pipelines like Unix tools. Single binary, supports piping and chaining.
3/7/2026
Show HN: EdgeDox – Offline document AI on Android using Qwen3.5-0.8B
Android app running Qwen3.5-0.8B locally for offline document AI without cloud upload. Supports PDF Q&A and document summarization.
3/7/2026
Should AI web agents skip sponsored/ad results by default?
Discussion of whether AI agents should skip sponsored ad results during automated web research to avoid accidental PPC click fraud.
3/7/2026
Show HN: CC Pocket – Control Claude Code/Codex from Your Phone
Mobile app to remotely control Claude Code and Codex CLI agents from iOS/Android via WebSocket bridge on Mac.
3/7/2026
Show HN: I built an AI agent that wrote a full novel in 10 minutes
Gollem: Go framework for production AI agents with type safety, structured output, multi-provider streaming, guardrails, and zero-allocation streaming.
3/7/2026
Google Always-On Memory Agent
Always-on memory agent using Google ADK and Gemini 3.1 Flash-Lite. Implements persistent 24/7 memory consolidation without vector databases or embeddings.
3/7/2026
Show HN: SurvivalIndex – which developer tools do AI agents choose?
SurvivalIndex project benchmarks what developer tools AI agents actually choose in practice. Finds Claude Code picks custom solutions over available tools in 60% of categories.
3/7/2026
Show HN: I accidentally caught an AI agent trying to poison my prod config
Pre-tool-use hook for Claude Code that logs all actions as tamper-evident audit trail, catching attempted production config poisoning.
3/7/2026
Show HN: Citepo-CLI, a lightweight CLI for creating blogs, build for AI agent
Lightweight CLI for creating and deploying MDX blogs. Designed with AI agents in mind for blog automation.
3/7/2026
Neolab and Emerging AI Lab Tracker
2026 directory of frontier AI labs and emerging research labs worldwide including AMI Labs, with contact and location info.
3/7/2026
"Clinejection" Turned an AI Bot into a Supply Chain Attack
Security disclosure of supply chain vulnerability in Cline AI coding tool via compromised issue triage bot. Real exploit of AI agent.
3/7/2026
Spark Runner: Easily Automate Front End Tests
Autonomous browser agent that decomposes test intentions into phases and executes frontend tests. AI-driven test automation tool.
3/7/2026
Show HN: Context-compact – Summarize agent context instead of truncating it
Technique to summarize AI agent context via compression instead of truncation. Preserves task continuity and memory for long-running agents.
3/7/2026
Show HN: Crypto data API where AI agents pay per request with USDC (x402)
Crypto data enrichment API on Base mainnet allowing AI agents to pay USDC per request for price data, trading signals, and market analysis.
3/7/2026
Natural Language AutoCoder Open SOurce
Open source AI assistant converting natural language to Python code. Features autonomous execution loop, persistent mistake memory, voice input, and internet integration.
3/7/2026
Show HN: Claude-consensus – Multi-model code review plugin for Claude Code
Claude-consensus: Multi-model code review plugin running GPT, Gemini, Grok in parallel for code analysis with consensus convergence.
3/7/2026
FASTEST LLM decode engine on Apple Silicon. 658 tok/s on M4-Max,beats MLX by 19%
MetalRT: LLM inference engine for Apple Silicon achieving 658 tok/s on M4 Max, 1.67x faster than llama.cpp in benchmarks.
3/7/2026
T3 Code: A Minimal Web GUI/Desktop App for Coding Agents
Minimal web GUI for AI coding agents. Currently supports Codex CLI with Claude Code support planned. Early stage.
3/7/2026
Teams have a context-sharing problem; TeamContext is our attempt
Git-native CLI tool for sharing LLM coding context between team members. Addresses context drift in multi-developer AI coding workflows.
3/7/2026
Show HN: I built an autonomous AI company that runs itself (22 cycles, $36)
Auto-Co: Autonomous AI company with specialized agent roles (CEO, CTO, engineer, etc.) running scheduled workflows, open source MIT license.
3/7/2026
LLM Doesn't Write Correct Code. It Writes Plausible Code
Analysis of LLM-generated code producing plausible but functionally incorrect implementations. SQLite lookup 20k times slower than expected.
3/7/2026
Show HN: Reflectt-node – tell Claude to install it, AI team in 5 min
Reflectt-node: Multi-agent coordination platform deployable via Claude/ChatGPT in 5 minutes with task board, chat, and review gates.
3/7/2026
Show HN: Git for your AI workflow - Version control for what Claude remembers
dullnote: Version control system for Claude's context/memory using MCP protocol, enabling persistent state across sessions and team sharing.
3/7/2026
Show HN: I wrote an LLM inference engine in pure Go – 48 tok/s zero dependencies
dlgo: Pure Go LLM inference engine loading GGUF models with no dependencies, supporting LLaMA, Qwen, Gemma, Phi models.
3/7/2026
I built the "Strava for Developers" because I'm tired of being a bar on a chart
Developer activity tracker with narrative context similar to Strava for fitness. Privacy-focused alternative to time trackers.
3/7/2026
Telemetry data analysis for any AI product
Atheon: Telemetry analytics platform for AI products tracking agent behavior, user intent, and failure modes without raw logs.
3/7/2026
An Agent ever delete something important to you
AI agent tool automating spreadsheet analysis, web research, form filling with browser automation and real-time monitoring.
3/7/2026
Show HN: CodeTrackr – open-source WakaTime alternative with real-time stats
CodeTrackr: Open-source privacy-first developer activity tracker with real-time dashboard, WakaTime-compatible API, Docker support.
3/7/2026
Show HN: OpenEHR-CLI – CLI and MCP server for working with openEHR artifacts
Open source CLI and MCP server for openEHR artifact management. Automates template validation and resource processing in scripts/CI. Integrates with Claude Desktop and Cursor via MCP.
3/7/2026
Show HN: Hatice – Autonomous Issue Orchestration with Claude Code Agent SDK
Hatice autonomously orchestrates GitHub/Linear issues by dispatching Claude Code agents to solve end-to-end. Built in TypeScript with multi-turn execution, retry logic, and observability.
3/7/2026
More Is Different for Intelligence
Essay on how software proliferation enables new kinds of processes beyond efficiency gains. Historical examples from business automation but limited technical depth.
3/7/2026
What if CLIs exposed machine-readable contracts for AI agents?
Sonde defines stable manifest contracts for AI-native CLI workflows with tooling to generate, validate, and score. Reference CLI implementation for standardized agent-CLI integration.
3/7/2026
Incrementally parsing LLM Markdown streams on server/client
Markdown parser with streaming support for incrementally parsing LLM outputs. Parses into typed node tree following CommonMark spec, enabling finalized block emission as input grows.
3/7/2026
Show HN: Cross-Claude MCP – Let multiple Claude instances talk to each other
MCP server enabling multiple Claude instances to communicate via shared message bus. Solves isolation problem in multi-session Claude workflows with messaging, channels, and data sharing.
3/7/2026
I'm 60 years old. Claude Code has ignited a passion again
Personal experience of developer reinvigorated by Claude Code agentic capabilities for programming.
3/6/2026
Agent Office – Slack for (OpenClaw Like) AI Agents
Agent Office provides Slack-like communication interface for AI agents similar to OpenClaw architecture.
3/6/2026
WebSocket+Huffman vs. SSE+JSON for streaming LLM tokens
Proof-of-concept comparing WebSocket+Huffman vs SSE+JSON for LLM token streaming. Achieves 30% latency improvement and 60% bandwidth savings through client-side token decoding.