9 topics covered

Listen to today's briefing
0:00--:--

Anthropic IPO Faces Investor Resistance

What happened: Reports say Anthropic's planned IPO is running into investor pushback over its $1.5 trillion valuation as the tech market sells off.

Key details:

  • Per The Verge, Anthropic filed its S-1 to go public the week before OpenAI's new funding news.

Why it matters: Investor resistance to the valuation would be a test of how the public listing is priced amid weaker tech stocks.

Practical takeaway: Monitor any revised valuation or timing changes for the Anthropic listing, which would indicate how the IPO is progressing.

Meta Muse Spaces Feature

What happened: Meta launched "spaces" in its Muse AI assistant, letting users build custom app-like experiences inside the app.

Key details:

  • Examples include a shopping hub, school dashboard, music tracker, and trivia game.
  • Users create a space by describing it.
  • Connected apps such as Gmail, Google Calendar, OpenTable, and Notion let the assistant update information like pricing and dates in the background.
  • Spaces can be kept in the sidebar, moved to the main tab, or scheduled to appear at certain times.
  • Meta is rolling out spaces to all Muse users and is exploring sharing and collaboration.
  • Meta also brought "Active Memory" and "Active Focus" from its Muse Wonder model into Muse and Meta AI.

Why it matters: Spaces extends Muse from a chat assistant toward persistent, continuously updated dashboards.

Practical takeaway: Muse users can start creating spaces now, and connecting outside apps is what enables automatic updates.

Ricursive Ripple-65B Open-Weight Model

What happened: Ricursive Intelligence released Ripple-65B Preview, a fully Apache 2.0 open-weight US language model focused on agentic coding.

Key details:

  • The model has 65 billion parameters.
  • The company says it achieves the top score among open-source US models on the TerminalBench 2.1 coding benchmark.
  • The company puts total R&D cost at $32 million.
  • On other benchmarks, Ripple falls behind Inkling 2 Small, the Thinking Machines open-weight model released alongside it.

Why it matters: It adds another US-developed open-weight option for coding agents, though it trails Inkling 2 Small on some benchmarks.

Practical takeaway: If you evaluate open coding models, test Ripple-65B Preview on your own agentic coding tasks, since its advantage is benchmark-specific.

OpenAI $11B Funding Round at $1.25T Valuation

What happened: OpenAI has secured $11 billion in fresh capital, the first tranche of a funding round that values the company at $1.25 trillion.

Key details:

  • The first $11 billion was reported by Bloomberg; The Verge notes the round could reach $20 billion in total.
  • OpenAI had earlier hoped to raise as much as $30 billion at a valuation of $1.4 trillion, according to The Verge, before tech stocks dropped and midterm results showed declining public support for AI companies and data centers.
  • OpenAI raised $122 billion from investors in March at a valuation of $852 billion, per The Verge.
  • The Decoder reports OpenAI is turning to wealthy private investors to fill out the round.

Why it matters: The lower valuation target and reliance on private investors, as The Decoder frames it, point to a more cautious AI capital market than the earlier round suggested.

Practical takeaway: Watch whether the full $20 billion round closes and at what valuation, since that will signal investor appetite ahead of Anthropic's planned listing.

Amazon Alexa for Shopping Text Ordering

What happened: Amazon added text messaging to Alexa for Shopping, letting customers buy items by texting the AI shopping assistant.

Key details:

  • Users can describe products or send a photo, and Alexa searches Amazon listings, texts back recommendations, and orders on their behalf.
  • The contact appears in phones as "Amazon Alexa."
  • Alexa says it remembers preferences over time and can track deals and prices.
  • The feature rolls out to a small group of US customers starting today, then to "millions more customers" in the coming weeks.
  • It builds on the Alexa for Shopping experience launched in May.

Why it matters: Texting extends agentic shopping into a channel users already use for shopping lists and messages.

Practical takeaway: US customers invited to the test can text Alexa to search, compare, and order items.

Protesters Breach Microsoft Amsterdam Data Center Site

What happened: Protesters broke into a construction site of a new Microsoft data center in Amsterdam, as part of a wave of European data center protests.

Key details:

  • Around 100 people reportedly entered the site, per Reuters as cited by The Verge.
  • Some protesters scaled a crane and placed a "fuck AI" banner; others unfurled a "Microsoft powers genocide" banner.
  • The Geef Tegengas group helped organize the protest against three 85-meter data center towers on the outskirts of Amsterdam.
  • Amsterdam banned new data centers until 2030 after a temporary stop announced at the end of 2023; the project predates the ban and is being built by Pure Data Centres, with Microsoft as sole tenant.
  • The same week, activists staged a mock Israeli military checkpoint outside a Microsoft data center in Dublin.

Why it matters: The action adds to growing opposition to data centers, now extending beyond the US to Europe.

Practical takeaway: Track whether Microsoft or local authorities respond to the Amsterdam site's permitting questions.

Google AI Blood Pressure Detection from Smartwatch

What happened: Google Research built an AI that detects signs of high blood pressure using only the pulse and motion sensors in a standard smartwatch, without a cuff.

Key details:

  • In a 196-person study, the system reached a 65.8 percent detection rate, which The Decoder says beat traditional in-office measurement.
  • The method is already live in the Google Health app.
  • It does not replace a formal diagnosis, according to The Decoder.

Why it matters: Cuff-free screening from wearables could make routine blood pressure checks more accessible, though the study is small.

Practical takeaway: Users of the Google Health app can check for the feature, but a formal diagnosis is still needed for any follow-up.

Ai2 ToyBlocks Robot Learning Framework

What happened: Ai2 introduced ToyBlocks, an open simulation framework for robot learning and evaluation, as the first "FoundationWorld."

Key details:

  • ToyBlocks is built around a robot arm and blocks on a tabletop, with tasks such as Stack, Unstack, Flip, and Rotate.
  • It automatically generates demonstrations without teleoperation, using reinforcement learning and motion planning in Nvidia's Isaac Lab.
  • The framework can run hundreds of planning jobs in parallel on a single GPU, reaching up to 24 times real-time performance.
  • Ai2 released code, data, models, and documentation.
  • A web interface lets users connect Claude or another frontier model to instruct the robot in natural language.
  • In its tests, the agent solved or correctly declared infeasible 63% of search-and-spell tasks, per the post.

Why it matters: A controlled, reproducible simulation setting gives researchers a way to isolate whether robot failures come from high-level planning or low-level control.

Practical takeaway: Robotics researchers can use the released code and models to test generalization and VLA behavior in a constrained environment.

METR Audit of Anthropic AI Acceleration

What happened: Nonprofit METR audited Anthropic's internal research data and concluded that Anthropic may have crossed its own AI acceleration warning threshold back in July.

Key details:

  • METR spent eleven weeks auditing Anthropic's internal research data.
  • METR concluded the pace of model development has roughly doubled.
  • About 80 percent of the speedup is driven by AI-assisted research, per METR.
  • Anthropic redacted the most informative evidence, so the public has to rely on METR's judgment "much more than would be ideal."
  • There are skeptical voices inside the company, according to The Decoder.

Why it matters: The audit suggests frontier labs' own acceleration thresholds may be crossed without full public visibility, since key evidence was redacted.

Practical takeaway: Read METR's findings with the redaction caveat in mind when assessing claims about Anthropic's development pace.