TSMC reports record profit and keeps its $60-64 billion spending plan

TSMC reported NT$763.06 billion in quarterly profit, up 68.7% from a year earlier. It kept its 2026 capital budget unchanged.

Read the full story →Listen to the edition

TSMC ↗TSMC ↗Focus Taiwan / Central News Agency ↗Associated Press ↗

Listen to this edition

Prefer audio? The briefing has chapters and a full transcript.

Follow on Apple Podcasts Follow on Spotify Transcript All episodes →

The briefing

The rest of the morning

4 more stories

Models and deploymentWarm

Meta releases weights for a model with 816 billion parameters

Meta released the weights for Muse Spark Open on October 15. The model has 816 billion parameters but uses 42 billion for each token. Its context window holds about one million tokens.

The Muse Model License adds conditions for companies that offer models as a service. A company must notify Meta if it and its affiliates earn more than US$20 million in gross revenue. The threshold applies to any consecutive 12-month period. The company then needs a separate written agreement before it can offer this model or its derivatives through that service.

Meta AI Research ↗Meta model repository ↗

Read article →

Developer securityWarm

GitHub releases AI scans, but they cannot block pull request merges

GitHub made AI Scan for pull requests generally available on October 15. It checks languages and frameworks that CodeQL does not cover today. The findings are advisory, so they cannot fail the code-scanning check or block merges through rulesets. AI Scan does not check pull requests from forks or Dependabot.

AI Scan requires code scanning and the relevant setting to be enabled, and it uses Code Security AI Credits. Private and internal organization repositories need Code Security or Advanced Security. Agentic autofix has separate eligibility on GitHub Enterprise Cloud. It needs administrator approval and a defined Copilot billing method, and uses AI Credits and Actions minutes. These are company-described tools. Passing a CodeQL recheck does not prove that all security problems are fixed.

GitHub Changelog ↗GitHub Changelog ↗

Read article →

Safety and governanceWarm

Anthropic gives six METR employees internal access and lets METR publish without approval

Anthropic said on October 15 that six METR employees now have access to its hardware and office space. They also have internal access similar to the access Anthropic gives its risk teams. METR can publish its assessments without Anthropic's approval.

Anthropic does not routinely share customer data, privileged information or details about its model architecture. The company granted this access voluntarily. This is not a government inspection or a finding that the models are safe.

Anthropic ↗METR ↗

Read article →

Medicine and societyWarm

952,008 clinicians asked OpenEvidence 271 million questions

A JAMA study published October 15 examined OpenEvidence use from January 1, 2024, through June 30, 2026. It counted 952,008 US clinicians, including 556,200 physicians, and 271 million queries. The other users included nurse practitioners, physician assistants and other licensed clinicians. Monthly active users reached 703,265 in June 2026. The study counted a user as active after the user submitted at least one question that month.

The observational study covers one platform. It does not show whether the answers were accurate or helped patients. The researchers could not tell whether clinicians asked questions during patient care, preparation or general learning.

JAMA ↗JAMA ↗Harvard Medical School ↗

Read article →

The watchlist

Signals that could change the read

Fourth-quarter revenue and the margin cost of new factories.

Independent tests, serving costs and the license terms for model hosts.

Useful GitHub repairs and METR disclosures about access limits.

Accuracy on treatment questions and measured patient outcomes.

Across the desks Warm

TSMC reports chip profits, while model users face separate costs, terms and checks.

ChipsBudget unchangedTSMC kept its 2026 capital budget at US$60-64 billion.

ModelsWeights releasedMeta released Muse Spark Open under a license with extra conditions for large businesses that host the model for customers.

SecurityFindings cannot block mergesGitHub says AI Scan findings cannot block pull request merges through rulesets.

MedicineQuestions measuredThe OpenEvidence study measured clinician questions, not answer accuracy or patient benefit.

Editorial direction, not a financial index. Each signal is tied to this edition’s reporting.

Edition validated · 5 stories · 13 unique sources

About HotTea & our sources →