Frontier Daily (Aug 1): 5 items from Latent Space, Simon Willison, OpenAI

Every day Laojin pulls the official channels and compiles the past 48 hours of AI news into a short brief: curation only, no commentary, with the original link and publish time attached to every item.

1. [AINews] not much happened today

This AI news roundup notes a post-training-only DeepSeek open-weights update that slightly advances the Pareto frontier, ending DeepSeek's year-long obscurity after its $70B pre-IPO fundraise.

Source: Latent Space · 2026-08-01 · Original

2. deepseek-ai/DeepSeek-V4-Flash-0731

DeepSeek's V4-Flash-0731, a 304B-parameter model with enhanced agentic capabilities, offers a strong cost-performance ratio and ranks ahead of MiniMax M3, though default reasoning produced mediocre results.

Source: Simon Willison's Weblog · 2026-07-31 · Original

3. Stateless MCP has recaptured my interest (and inspired mcp-explorer and datasette-mcp)

The stateless MCP 2.0 spec reignites the author's interest in the protocol, as it reduces implementation complexity and makes tools easier to audit and control, leading to three new builds including mcp-explorer.

Source: Simon Willison's Weblog · 2026-07-31 · Original

4. Building abundant intelligence

OpenAI describes a full-stack approach to making advanced AI more capable, affordable, and widely useful.

Source: OpenAI Blog · 2026-07-31 · Original

5. Advancing responsible AI across Europe

OpenAI outlines how its safety, security, transparency, and provenance practices support responsible AI governance in Europe as the EU AI Act advances.

Source: OpenAI Blog · 2026-07-31 · Original

---

_Sources: official blogs of OpenAI, Google DeepMind and Qwen; official RSS of Simon Willison, Latent Space, Import AI and One Useful Thing; plus change monitoring of anthropic.com. Mistral / Meta AI have no stable public feed; X sources are curated by hand. This is a source digest — no commentary from Laojin._