This week we reviewed 252 articles from 22 sources.
Apple sues OpenAI for allegedly stealing hardware secrets, two senior OpenAI leaders depart, and SpaceX AI ships Grok 4.5 at $2/M tokens. The community builds practical tools for pruning long-context costs and orchestrating agent swarms.
Official Highlights
Apple alleges OpenAI senior leadership encouraged poached employees to bring confidential presentations, secret prototypes, and supplier details to the company. Read more
OpenAI launches GPT-5.6, its latest family of models, promising improvements across cybersecurity and other domains. Read more
Fidji Simo steps down as OpenAI's CEO of AGI Deployment after extended medical leave; she stays on as part-time adviser amid a concurrent Head of Safety departure. Read more
SpaceX AI ships Grok 4.5, a Cursor-trained coding model serving at 80 TPS and ranking #1 on Harvey's Legal Agent Benchmark. Read more
Meta Superintelligence Labs releases Muse Spark 1.1, a multimodal reasoning model with 1M-token context built for agentic tasks on the Meta Model API. Read more
From the Community
Eivind Kjosbakken's repo shows how to avoid deadlocks when 100 agents share a Claude context window—worth stealing if you're hitting rate limits. Read more
Emmimal P Alexander built a prompt-pruning layer that drops redundant tokens from long contexts, cutting inference costs without degrading output quality. Read more
Omer Rosenbaum blames attention dilution in 100K-token contexts for frontier hallucinations and ships a Colab with the fix—test it on your own prompts. Read more
Bala Priya C maps agent memory types to tasks through a decision tree, showing when to reach for episodic, semantic, or procedural stores. Read more
The running list of major tech layoffs in 2026 where employers cited AI as a factor, tracked in reverse chronological order. Read more
A Coding Guide to NVIDIA's Tile-Based GPU Programming walks through TileGym, cuTile, Triton kernels, and Flash Attention on a Colab that runs across hardware tiers. Read more
Featured This Week
Emmimal P Alexander's "Long context isn't free" attacks a production problem most teams ignore: LLMs degrade not from forgetting but from remembering too much. The post details a pruning layer that scores attention weights and drops redundant tokens, with a shipped Colab containing the exact heuristics and A/B results. For engineers piping documents, logs, or multi-turn conversations into production systems, this moves long context from demo trick to cost-controlled architecture. Read more
Send comments and story tips to tips@planet-ai.net. If you found this useful, share it with a colleague.
This Week in AI
Additional stories worth scanning. Title only — click through for the full piece.
Scientists’ Side Hustle? Using AI and Quantum Computing to Generate New Peptides — WIRED · 2026-07-12
OpenAI bets on families as ChatGPT goes deeper into households — TechCrunch · 2026-07-11
Meta removes controversial AI feature on Instagram after backlash — TechCrunch · 2026-07-10
Lovable reportedly in talks to double its valuation to $13.2B — TechCrunch · 2026-07-08
Meta rolls out Muse, a new AI image generator — TechCrunch · 2026-07-07
The ‘first’ AI-run ransomware attack still needed a human — TechCrunch · 2026-07-06
Amazon will stop accepting new customers for Mechanical Turk — TechCrunch · 2026-07-05
Messi and Ronaldo Are Building Tech Portfolios. Mo Salah Is Playing a Different Game — WIRED · 2026-07-08
Meta Now Lets Anyone Use Your Instagram Photos in AI Images—Unless You Opt Out — WIRED · 2026-07-07
US investors will soon get access to SK Hynix, another memory maker riding the AI boom — TechCrunch · 2026-07-06
Apple sues OpenAI over alleged trade secret theft — TechCrunch · 2026-07-10
Google’s deepfake detector system used to debunk McConnell hoax pic — TechCrunch · 2026-07-08
OpenAI’s Chief Futurist Is Leaving the Company — WIRED · 2026-07-07
Vercel CEO Guillermo Rauch on the fight to split off models from agents — TechCrunch · 2026-07-06
Open source AI matters more than ever, according to Hugging Face’s Clem Delangue — TechCrunch · 2026-07-10
An AI agent startup just let its agent run its $100 million fundraise — TechCrunch · 2026-07-09
I Built a Self-Improving AI, and So Can You — WIRED · 2026-07-08
Why the rise of open source AI isn’t hurting Anthropic … yet — TechCrunch · 2026-07-07
You can now customize Siri’s pace and expressivity in the latest iOS 27 beta — TechCrunch · 2026-07-06
SK Hynix raises $26.5B in the biggest foreign IPO in US history, is urged to build new US fabs — TechCrunch · 2026-07-10
OpenAI is shutting down Atlas, but its AI browser ambitions are still growing — TechCrunch · 2026-07-09
SpaceX AI releases Grok 4.5, which Elon describes as an ‘Opus-class model’ — TechCrunch · 2026-07-08
Microsoft joins AI cost-cutting trend by relying more on its own models — TechCrunch · 2026-07-07
Hugging Face’s CEO on why companies are done renting their AI — TechCrunch · 2026-07-10
Elon Musk praises Mythos/Fable, promises not to ‘cut off’ Anthropic — TechCrunch · 2026-07-09
This newsletter supports planet-ai.net, a curated aggregator for AI tutorials and official updates. Curated by Keith Larson.
No comments:
Post a Comment