📰
subtl daily briefing
Latest Developments
⚡
Everything else in the news today
Eric shares that a strong 'why' matters more than time management tricks when balancing a full-time job and side projects.→
Metricool sponsors the newsletter, offering Superpath readers a free plan plus 30 days of paid toolkit with code SUPERPATH.→
Superpath is hosting an IRL Brooklyn Content Marketing Meetup on Sep 29 and released a new Content, Briefly podcast episode with Jimmy.→
Inter Milan cut social content by a third but grew followers 20% YoY, impressions 43%, and video views 33%.→
82% of global consumers use AI in 2026, but AI only drives 3.7% of shopping inspiration, per VML.→
ChatGPT and Google Gemini lead AI platform usage; brands must ensure visibility in AI answers.→
DeepLearning.AI launched a new short course, Building AI Assistants with On-Device Memory, built with Qdrant and taught by Dylan Couzon.→
The course teaches storing, searching, filtering, and forgetting text and image memories locally so nothing leaves the device.→
It also includes a new AI coding lab in the DeepLearning.AI mobile app for practicing with an AI coding agent.→
Muse hit #1 on both the U.S. App Store and Google Play within 12 days, beating ChatGPT's launch momentum.→
Meta shares surged 11.4% to $741 after JPMorgan, Wells Fargo, and KeyBanc raised price targets, and Meta One subscriptions reached 15M.→
Amazon blocked Muse from its store while Shopify enabled agentic checkout with Shop Pay, highlighting who controls AI agent access.→
Anthropic released Claude Opus 5.5 at 40% lower cost and 30% faster output, with prices at $4/$20 per million tokens and four breaking changes for existing integrations.→
OpenAI shipped GPT-6 Sol ($2/$10 per million tokens) and GPT-6 Luna ($0.10/$0.50), both with 1M context windows and 90% cached-input discounts.→
A study of 69,632 medical preprints found em-dash use nearly tripled after ChatGPT launched, a signal of AI's footprint on scientific writing.→
Rand Fishkin published a new SparkToro blog post arguing that algorithms are trained by user behavior, using the iced coffee debate as an example.→
The post suggests that public controversies like iced coffee in job interviews shape how algorithms are trained.→
The iced coffee controversy is used to illustrate how users train algorithms through their reactions.→
Microsoft launched a new AI-ready datacenter in Hyderabad, India, with zero-water cooling and a $20.5 billion investment.→
Meta's macOS AI assistant Muse had a serious 0-day vulnerability allowing local apps to steal tokens and access sensitive data.→