3. Show HN: cdai β the AI-enhanced cd command (github.com)
Contribute to franzenzenhofer/cdai development by creating an account on GitHub.
4. Show HN: 2x-4x cheaper GLM 5.3 for coding and research (www.coralbricks.ai)
Same model, same prompt β multiple times the tokens per second, with near-zero rate limits. Inference built for long-running headless agents, so the runs that used to queue now finish on time.
5. Show HN: Slotstream, run Qwen3.8-Flash-Next 4-bit on a low-memory Mac (github.com)
Run Qwen3.8-Flash-Next (125B MoE, 104 GB at 4-bit) on Macs with a fraction of that RAM by streaming experts from SSD. MLX + Swift, Ollama-compatible API. - carloslfu/slotstream
6. Show HN: WOML β an HTML-like language for building workflows (woml.org)
WOML is an HTML-inspired language and self-hosted runtime for readable, durable workflow automation with real JavaScript.
8. Show HN: VajraClaw β Deterministic <1Β΅s execution guardrail for AI agents (github.com)
β‘ DROSβ’ VajraClaw (Hacker Edition): Free Multi-Agent Runtime Security & Docker Governance Gateway with W3C DID & <1ΞΌs Enforcement - Top-Celestial-Company-Ltd/DROS-VajraClaw-Hacker
9. Show HN: A Pi extension to monitor your quota for OpenCode GO et CommandCode (github.com)
Pi extension β shows subscription quota usage (%) in the status bar for Command Code and OpenCode Go, next to the TPS readout - Anhydrite/pi-quota-monitoring
10. Show HN: Sonde, a local code graph for AI agents that refuses to guess (github.com)
Local code-context engine for AI coding agents: structural answers within a token budget. - anishmoncivarghese/sonde
13. Show HN: Type (type.thingg.co)
An Android keyboard with on-device typo correction, neural next-word prediction, and personal learning. Nothing you type leaves the device.
15. Show HN: Card Shop Directory β Find trading card shops by state, city, and game (cardshopdir.com)
Browse 7,700+ trading card shops across all 50 US states. Find local game stores for Pokemon, MTG, Yu-Gi-Oh!, and more.
16. Show HN: What Apple's OS updates silently change in the on-device AI model (umer9538.github.io)
One frozen prompt suite, captured per OS build, diffed. Superseded builds can never be measured again β every capture here is history that cannot be re-taken.