TLDR AI 2025-08-14
xAI co-founder leaves 👋, Gemini memory 🧠, Gemini Deep Research API 👨💻
AI traffic is up 527%. SEO is being rewritten (6 minute read)
AI-driven web traffic surged 527% from January to May, with LLMs like ChatGPT and Perplexity reshaping user engagement. Legal, Finance, and Health industries lead in AI-sourced sessions, indicating a shift towards AI for complex inquiries. SEO must adapt to prioritize clarity and relevance for AI selection, or risk falling behind in user discovery.
gpt-oss-120B performance varies dramatically by provider (10 minute read)
Artificial Analysis benchmarked OpenAI's gpt-oss-120B across 11 API providers and found striking performance differences despite hosting the same model. AIME math scores ranged from 78.3% on Azure to 93.3% on Cerebras — a 15-point gap that suggests provider infrastructure significantly impacts model capabilities. The major cloud providers (Azure, Amazon Bedrock, and Nebius Base) consistently lag behind smaller, specialized AI inference startups like Cerebras, Groq, and Fireworks.
👨💻
Engineering & Research
The AI Runtime driving top consumer AI apps from prototype to 1M users in <30 days is now available publicly (Sponsor)
- Adaptive Graphs auto-scale workloads from 10 to 10M users. C++ speed, any model.
- Automated MLOps handles ops, telemetry, and optimizations automatically.
- Live Experiments run with one-click, no code changes.
Try Inworld Runtime. Easily integrates with your existing stack and all providers. Free through August.
Void (GitHub Repo)
An open‑source, VS Code–derived AI code editor offering direct connections to any LLM (local or cloud), agent modes, change checkpoints, and full data privacy—with no backend middleman.
Google is working on the Deep Research API (1 minute read)
The Gemini Deep Research team is working on the Deep Research API. This will be a big upgrade for builders. Deep Research can automatically browse hundreds of websites, think through findings, and create insightful multi-page reports in minutes. It is designed to tackle complex research tasks by breaking them down, exploring the web to find answers, and synthesizing findings into comprehensive results.
iMessage MCP: Give Your LLM Access to Your Messages (6 minute read)
@wyattjoh/imessage and @wyattjoh/imessage-mcp on JSR are open-source packages that give LLMs read-only access to users' iMessage database on macOS. They enable natural language queries for messages, with all processing done locally on users' machines. This post walks users through the installation process and getting started with the tools.
Get the most interesting AI stories and breakthroughs delivered in a free daily email.
Join 920,000 readers for
one daily email