Shipped.
Cover line

The week Anthropic filed its S-1, set a public timeline for Mythos, deprecated Opus 4.1, and shipped six Claude Code builds.

The Open

Monday, June 1. Anthropic submitted a confidential draft registration statement on Form S-1 to the U.S. Securities and Exchange Commission. (anthropic.com) The direction: public markets. The timeline: as early as October 2026.

The rest of the week did not pause for the filing. Thirty-four hours later, Glasswing expanded to approximately 200 organizations across 15-plus countries, with Mythos-class models “in the coming weeks” buried near the bottom of the announcement. Wednesday: a year of AI threat data mapped to MITRE ATT&CK and contributed to Verizon’s DBIR, and the Claude Partner Network got formal tiers. Thursday and Friday: four more Claude Code builds, the last two focused squarely on enterprise fleet management. Friday afternoon: TypeScript and Python SDKs coordinated a release with client middleware and Opus 4.1 formally deprecated across every platform.

There is a thread connecting the product sprint to the filing. The 10,000 high-severity flaws found by Glasswing, the threat analysis mapped to MITRE ATT&CK, the version pinning in Claude Code for enterprise fleets: these are not incidental. They are the evidence the S-1 prospectus will eventually describe.

The Lead Story

$965 billion, confidential

The S-1 mechanics are these. Anthropic submitted a confidential draft on Form S-1 to the SEC on June 1, 2026. Confidential status means the document’s risk factors, revenue details, margins, and growth trajectory stay private until the company elects to disclose them. What is public: the direction and the scale. (anthropic.com)

What had already leaked into reporting: the annualized revenue run rate has crossed $47 billion. (CNBC, NPR) That number deserves a frame. Salesforce, which built its enterprise platform over two decades, reported $34.9 billion for all of fiscal 2024. Anthropic’s current run rate is 35 percent above that. The velocity: $9 billion at year-end 2025, $14 billion in February, $30 billion in April, $47 billion now. (Fortune) Four data points in five months.

The line is nearly vertical.

Target listing: as early as October 2026. Wilson Sonsini, the firm that handled Google’s 2004 IPO, is reportedly engaged as legal counsel. No share count. No price range. Both wait on market conditions and the eventual public prospectus.

What the filing will need to address: the cost of safety. Anthropic’s Responsible Scaling Policy, alignment research compute, the Glasswing infrastructure, and the interpretability evaluations required to clear Mythos for public release are all real expenditures. None appear in the run-rate number. Public markets will want to know how they flow through to margins.

What the filing will not need to justify, at this valuation: the market’s belief in the product. $47 billion in annualized revenue settled that argument. The S-1, when it becomes public, is the document that explains the gap between the safety story and the growth story. It will be the read of the year.

Also Shipped
Glasswing expanded, and Mythos got a timeline.
Anthropic added 150 organizations across 15-plus countries to Project Glasswing, bringing total partners to approximately 200. (anthropic.com) New sectors: power, water, healthcare, communications, hardware. Samsung and NATO confirmed as additions, per Financial Times reporting. Partners have collectively found more than 10,000 high- or critical-severity security flaws in widely deployed systems since launch, per CNBC. The access program is not ending. It is becoming the on-ramp to a public product: Mythos-class models will be available to all customers “in the coming weeks,” once additional safeguards are ready.
A year of AI threats, mapped.
Wednesday, Anthropic published an analysis of 832 accounts banned for malicious cyber activity between March 2025 and March 2026, mapped to MITRE ATT&CK. (anthropic.com) AI use concentrates at later-stage, higher-complexity attack operations, not reconnaissance. The activity is becoming more autonomous. Anthropic contributed the analysis to Verizon’s 2026 Data Breach Investigations Report, placing it in the industry’s canonical threat taxonomy. The timing, one day after expanding Glasswing, was not accidental.
The partner network got tiers.
Wednesday, the Claude Partner Network formalized three service tiers with a public portal. (anthropic.com) Select: 10 certified practitioners, 2 production deployments. Preferred: 100 practitioners, 15 active customers. Global Premier: 1,000 practitioners, 100 customers across 3-plus regions, executive-backed joint business plan. The Claude Partner Hub gives partners a daily-refreshed view of where they stand. Since March 2026: 40,000-plus firms applied, 10,000-plus consultants certified.
Claude Code: six builds, one arc.
Six Claude Code versions shipped Tuesday through Friday. The early-week builds (v2.1.160, v2.1.161) tightened security: explicit prompts before writing shell startup files, OTEL labels on metric datapoints, MCP credential redaction. Wednesday’s v2.1.162 added waitingFor to claude agents --json, so automation knows what a blocked session is actually blocked on. Thursday’s v2.1.163 added managed version pinning via requiredMinimumVersion and requiredMaximumVersion, so organizations can lock Claude Code to an approved version range across developer fleets and CI pipelines. Friday’s v2.1.166 added fallbackModel configuration: up to three alternative models tried in order when the primary is overloaded. (GitHub) The arc across the week: the team is building for fleets.
SDK day: middleware and a deprecation.
Friday, Anthropic coordinated a release across all TypeScript platform SDKs and the Python SDK. @anthropic-ai/sdk v0.101.0 adds client middleware to the HTTP stack, mirrored across the Vertex, Foundry, Bedrock, and AWS platform wrappers. Claude Opus 4.1 is formally deprecated across all five packages. Python SDK anthropic v0.106.0 carries the same deprecation plus a schema transform fix ($defs preserved when schema root is a $ref) and a Foundry client fix. Security disclosures now route to HackerOne.
Service disruption (resolved).
Elevated error rates across Claude AI, Claude Console, Claude API, and Claude Code were reported starting at 2:19 AM ET on June 2. Root cause identified and resolved same day. No public post-mortem at time of this compilation.
Quiet on the Wire
No new model releases in the window. API, Claude Apps, and system-prompt release notes documentation pages returned HTTP 403 throughout the week’s sweeps; those areas cannot be confirmed for this record. The S-1 is confidential: no follow-on public disclosures from the filing in the same week it was filed.
The Close

Anthropic filed for its IPO on Monday and shipped six Claude Code builds by Friday.

The $47 billion run rate and the version pinning for enterprise fleets are the same story.

The prospectus, when it arrives, will be the read of the year.

Release Log

A 1:1 mirror of every Anthropic release for the week of 2026-06-01. Sources: daily pages 2026-06-01 through 2026-06-05.

Date Area Release One-liner
2026-06-01 News Anthropic confidentially submits draft S-1 to SEC IPO process begins; $965B post-money valuation, $47B annualized run rate, October 2026 target.
2026-06-02 News Project Glasswing expanded to 150 new organizations Samsung, NATO join; 10,000+ critical flaws found; Mythos GA "in coming weeks."
2026-06-02 News Service disruption across Claude products (resolved) Elevated error rates on Claude AI, Console, API, Code from 2:19 AM ET; resolved same day.
2026-06-02 Code Claude Code v2.1.160 Security prompts for shell startup files; build-tool config approval; “workflow” keyword trigger config.
2026-06-02 Code Claude Code v2.1.161 OTEL labels on metrics; agent progress rows; MCP credential redaction; parallel tool call isolation.
2026-06-03 Research AI-enabled cyber threats mapped to MITRE ATT&CK 832 banned accounts analyzed Mar 2025-Mar 2026; contributed to Verizon 2026 DBIR.
2026-06-03 News Claude Partner Network Services Track and Partner Hub Three tiers (Select, Preferred, Global Premier); 40k+ applicants, 10k+ certified consultants.
2026-06-03 Code Claude Code v2.1.162 waitingFor in agents JSON; dedicated Grep/Glob; Remote Control footer pill; slash-command fill.
2026-06-04 Code Claude Code v2.1.163 Managed version pinning (requiredMinimumVersion, requiredMaximumVersion) for fleet and CI lock-in.
2026-06-05 Code Claude Code v2.1.165 Bug fixes; no user-facing changes.
2026-06-05 Code Claude Code v2.1.166 fallbackModel config; glob deny rules; cross-session security hardening; JetBrains 2026.1 fix.
2026-06-05 SDK-TS @anthropic-ai/sdk v0.101.0 Client middleware; streaming stop_details fix; JSON number parsing fix.
2026-06-05 SDK-TS vertex/foundry/bedrock/aws-sdk updates Opus 4.1 deprecated across all four TypeScript platform wrappers; client middleware added.
2026-06-05 SDK-PY anthropic v0.106.0 Opus 4.1 deprecated; Foundry copy()/with_options() fix; schema $defs/$ref transform fix.
Stay on the frontier

Get Shipped. in your inbox.

Daily digest at 9 PM ET. Weekly magazine every Friday morning. Six labs, one feed. No spam, one-click unsubscribe.