PostHog /posthog
PostHog is an open-source platform for building self-driving products, offering developer tools
→ View original sourcememvid /memvid
Memvid is a memory layer designed for AI agents, offering serverless, single-file implementation to replace complex RAG pipelines. It enables instant retrieval and long-term memory for agents, streamlining memory managem…
→ View original sourceAt least 105 past YC founders have worked at OpenAI and Anthropic
(No description available)
→ View original source
I ran a 7B Mixture-of-Experts LLM in a language I built — token-identical to fp32
I've been building machin (MFL), a machine-first language, and using it to write an LLM inference engine with zero dependencies — no PyTorch, no llama.cpp, no BLAS, no Python at runtime. Just a static
→ View original source![Kimi K3 Release Video [Made with Kimi K3]](https://external-preview.redd.it/MnJzMTRhc2t0bWRoMWHTmT_WvVFt6Y5TSWurs0nBBOCVV4no67AbAH_kt0mV.png?width=640&crop=smart&auto=webp&s=8cd07e9632e7c8e136f799d729428589a58c1807)
Kimi K3 Release Video [Made with Kimi K3]
The post announces a release video for the Kimi K3 model, demonstrating its capabilities for creative tasks and noting that it feels superior to the earlier GLM 5.2 model. The author plans to add more benchmark examples,…
→ View original sourceHKUDS /nanobot
HKUDS presents nanobot, a lightweight open‑source AI agent designed to integrate with tools, chats, and workflows. It is hosted on GitHub under the HKUDS organization and aims to enhance automation and interaction for de…
→ View original source
SPROG-9M: how far a 9-million-parameter, LLM-free model gets on grade-school math
The SPROG-9M model is a 9-million-parameter, LLM-free architecture designed to perform grade-school mathematics. The research explores the performance limits of such small-scale models on mathematical reasoning tasks wit…
→ View original source
Can LLMs Perform Deep Technical Comprehension of Computer Architecture Papers
A recent arXiv study examines whether large language models (LLMs) can achieve deep technical comprehension of computer architecture research
→ View original source
Speculative Decoding: Faster LLM Inference Without Quality Loss
Speculative decoding accelerates large language model inference by drafting several tokens cheaply and validating them with a single target-model pass. A rejection-sampling step ensures the resulting distribution matches…
→ View original source
Which llama.cpp is FASTEST for Qwen3.6 MTP + Vision? failed requests.
A benchmark of 17 different llama.cpp commits was conducted to determine the fastest version for Qwen3.6 MTP + Vision, totaling 153 successful samples. The commit 57fe1f07c3b6 (b9620), published on June 13, 2026, emerged…
→ View original source
Building a multilingual voice agent: lessons from FR/EN/DE: field controls that hold
The article stresses that multilingual voice agent development must avoid filler content and instead present a clear method, explicit limits, and verifiable sources to maintain credibility in Voice AI research and open‑s…
→ View original source