📰
subtl daily briefing
Latest Developments
⚡
Everything else in the news today
Solana ETFs pulled in a record $188M in one week and validators voted to halve new SOL issuance, but John remains cautious short-term with the macro index still RISK OFF.→
Alpenglow went live on testnet, cutting Solana's finality from ~12.8 seconds to ~150ms, but no mainnet date is set.→
Robinhood Chain has overtaken Solana in tokenized stock trading at ~39% market share, while Solana clawed back to ~35%.→
Marketing Brew Summit in-person tickets are sold out; the livestream is available at 50% off with code NOFLIGHTS.→
Livestream ticket holders get on-demand access to session recordings a few weeks after the event.→
Sponsors include Profound, Bloomreach, Awin, Hammermill, Optimizely, and Yahoo.→
Cakes Body has generated $100M in annual gross revenue for two straight years via founder-led marketing on TikTok Shop.→
FTC Green Guides are outdated, and experts like Clorox's La Toya Sutton advise avoiding vague terms like 'sustainable.'→
Frontify is hosting an AI-branding event on Oct. 6 with leaders from Google DeepMind, MANN+HUMMEL, and Peloton.→
DeepLearning.AI launched a 4-course Data Engineering Professional Certificate to prepare learners for data engineering roles.→
The program covers the full data lifecycle: generating, ingesting, storing, transforming, and serving data.→
Learners will build batch and streaming pipelines and implement modern data architectures with open-source tools.→
Muse hits 500k downloads and 250k active users in week one; Matter integrates it into new glasses and a 'Charm' puck.→
OpenAI discovers its testing agents scraped an Australian health site, exposing engineering and management failures.→
Anthropic used Claude to discover a novel enzyme system, demonstrating the power of parallel model swarms.→
Salesforce now meters third-party agents via MCP/API with Flex Credit charges, while HubSpot raises prices for its own Breeze agents but leaves its MCP server free.→
Larridin is SaaStr's AI App of the Week, helping companies track what the median engineer's $213/week in AI coding tokens actually buys.→
SaaStr argues great VPs of Sales never hide from a miss, and the two fastest signs of a mis-hire are no one great joins in 30 days and nothing improves in one sales cycle.→
Claude Sonnet 5.5 hits 70.6% on Terminal-Bench 4.0, beating Opus 5.5 (66.4%) while being 30% faster and up to 30% cheaper per task.→
OpenAI patched a GPT-6 image understanding bug that was silently degrading API, Codex, and computer use results.→