[ The Morning Build ]
3 min read5 storiesAIBig Tech

The Morning Build for September 24, 2026: Anthropic's Claude Finds a CRISPR-like System, Muse Expands, and Gemini 3.8 TTS Ships

Today’s stories connect where models meet real-world systems: Anthropic used Claude agents to discover a previously uncharacterized enzyme system associated with repeat arrays; Meta pushed Muse deeper into devices, apps, and commerce; YouTube and YouTube Music rolled AI-curated feeds and conversational music tools; and Google shipped Gemini 3.8 text-to-speech models.

Anthropic’s Claude autonomously discovers array-associated reverse transcriptases (ART) linked to CRISPR-like repeat arrays

  • What happened: Anthropic reports that Claude agents scanned ~210 million tokens across DNA databases, collecting over 200,000 reverse transcriptases, narrowing 3,500 candidates to 20 high-priority systems, and leading to the identification and lab follow-up of a previously uncharacterized enzyme system they call ART, found mainly in bacteriophages and composed of an RT, a partner gene, and a tandem repeat array.
  • Why it matters: This is an example of an LLM-driven pipeline producing testable biological hypotheses at scale: Claude produced candidate reports that materially reduced time from weeks or months of expert analysis to an automated shortlist suitable for experimental validation, and initial lab work shows the array is expressed as short RNAs.
  • Outlook: Anthropic will publish further experimental results and updates in its technical report and pre-print as follow-up experiments proceed; the company said experiments are ongoing and the pre-print is available now.

Sources: anthropic.com · the-decoder.com

Meta expands Muse with realtime avatars, Mac control, smart-glasses integration, and commerce connectors

  • What happened: At Connect, Meta announced Muse Realtime Avatar for live video-chat avatars, Mac control so Muse can operate desktop apps, smart-glasses integration that lets Muse respond via a wake word, and new commerce and app connectors including Shopify, Stripe, PayPal, Best Buy, Gap, Sephora, Walmart, Wayfair, Expedia, and soon Instacart; Meta also opened a developer connector program and received over 1,500 applications in under a week.
  • Why it matters: Meta is extending agent capabilities into hardware, desktop automation, and payments connectors, which changes integration points for developers: Muse Spark powers agentic workflows across devices, and connectors plus commerce partners imply agent-mediated transactions and third-party integration surfaces to build against.
  • Outlook: Smart-glasses integration and several new Muse features are slated to launch in “the coming months,” and Meta said Muse email accounts and additional connectors will arrive “soon,” making these rollout windows the next public milestones.

Sources: techcrunch.com · the-decoder.com · techcrunch.com

YouTube launches AI-powered custom feeds so users can describe and pin personalized recommendation channels

  • What happened: YouTube introduced ‘custom feeds,’ a prompt-driven feature powered by Google’s Gemini model that lets users describe the videos they want recommended; created feeds are pinned in a dedicated tab and will not replace the main recommendation feed.
  • Why it matters: Developers and product teams should expect user-level algorithmic controls driven by large models: Gemini composes feed rules from natural-language prompts and YouTube will surface multiple pinned feeds, changing how recommendation logic and personalization are exposed to end users.
  • Outlook: Support for creating multiple custom feeds will roll out on web and mobile starting next month, which is the first concrete product milestone for developers and users to check.

Sources: techcrunch.com

YouTube Music adds ‘Ask Music’ conversational queries and ‘Your Podcast Lineup’ personalized audio guide

  • What happened: YouTube Music announced Ask Music, a conversational tool that uses natural language to build listening queues and answer music- and podcast-related questions across its catalog of over 300 million songs, and Your Podcast Lineup, a weekly AI-generated spoken preview of recommended podcasts that can move into suggested episodes.
  • Why it matters: Embedding conversational agents directly into a streaming music product shifts discovery from keyword search to intent-oriented prompts, and YouTube Music will deliver context-aware queues and weekly generated audio summaries to paid subscribers, changing product expectations for music and podcast discovery.
  • Outlook: Both Ask Music and Your Podcast Lineup are scheduled to arrive soon for YouTube Music and YouTube Premium subscribers worldwide, making the subscriber rollout window the next public milestone.

Sources: techcrunch.com · the-decoder.com · techcrunch.com

Google releases Gemini 3.8 Flash and Flash-Lite TTS for expressive, prompt-driven voice design

  • What happened: Google announced Gemini 3.8 Flash TTS for granular creative voice design and Gemini 3.8 Flash-Lite TTS for high-volume, cost-efficient speech; both support line-by-line performance direction, long-form generation, dual-speaker staging, over 100 languages, 2,000+ production voices, 30-second voice replication with consent verification, SynthID watermarking, and availability in Gemini API and Google AI Studio starting today, with enterprise API access coming soon.
  • Why it matters: Developers gain prompt-driven voice synthesis with fine-grained control, long-form stability, and built-in provenance via SynthID and consent verification, enabling production dubbing, voice agents, and multi-speaker audio pipelines at scale through Gemini API and Google AI Studio integrations.
  • Outlook: Gemini 3.8 Flash TTS and Flash-Lite are rolling out today to developers in the Gemini API and Google AI Studio, with Gemini Enterprise API access listed as coming soon and enterprise availability the next milestone to watch.

Sources: blog.google · the-decoder.com