---
Title: Anthropic's IPO filing puts numbers on Claude Code's economics
Author: Erik Benjaminson
Published: 2026-10-27T00:08:00.321Z
Description: Anthropic filed its S-1 on October 26. The same week two Claude Code builders left, users reported smaller limits, and Duet opened a waitlist.
URL: https://sapienttech.dev/pulse/2026-10-27-anthropic-s-ipo-filing-puts-numbers-on-claude-code-s-economi
---

<p class="text-sm opacity-80">Coverage: the past 7 days. Searched “Claude Code”, “Anthropic”, “Claude Code Anthropic departures”, “Claude usage limits”, “Claude Agent SDK” on News, Hacker News, Reddit, GitHub, and X.</p>

<p>Anthropic filed its S-1 on October 26. The same week two Claude Code builders left, users reported smaller limits, Claude refused reverse engineering, and Duet opened a waitlist.</p>

<p>Anthropic filed its S-1 on October 26 to list on Nasdaq as ANT. First-half revenue was $16.3 billion, and the $80 billion net loss includes a $72 billion non-cash charge. The S-1 is Anthropic's first public account of its revenue, costs, and customer concentration, including how its heaviest Claude Code users spend.</p>

<details>
<summary>Evidence and sources</summary>
<div class="my-3">Net revenue retention is about 900%. 11,900 customers pay over $100,000, 1,900 over $1 million, and about 250 over $10 million.</div>
<div class="my-3">Opus 5.5 cut customer cost per task about 40% against Opus 5. Compute fell from 201% of revenue to 55%.</div>
<div class="my-3">Claude Code's top enterprise users outspend other Claude users 20 to 1, per The New Stack's reading of the filing.</div>
<div class="my-3"><a href="https://www.saastr.com/5-interesting-learnings-from-anthropics-s-1-at-16-3-billion-in-half-year-revenue-900-nrr-6-6m-in-revenue-per-employee-and-a-10-adjusted-operating-margin">SaaStr on the S-1: revenue, net revenue retention, customer tiers, and compute ratio</a> <span class="opacity-70 text-sm">(News, Oct 26)</span> · <a href="https://thenewstack.io/claude-codes-enterprise-20-to-1">The New Stack: top enterprise Claude Code users outspend other Claude users 20 to 1</a> <span class="opacity-70 text-sm">(News, Oct 26)</span> · <a href="https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm">Anthropic's S-1 filing on the SEC site</a> <span class="opacity-70 text-sm">(via Hacker News, Oct 26)</span> · <a href="https://www.reddit.com/r/technology/comments/1xb7op7/anthropic_logged_net_loss_of_over_80_billion_in/">The $80 billion net loss, posted six days before the filing</a> <span class="opacity-70 text-sm">(Reddit, Oct 20)</span> · <a href="https://x.com/AnthropicAI/status/2114705042817110293">Anthropic's S-1 announcement</a> <span class="opacity-70 text-sm">(X, Oct 26)</span></div>
</details>

## Anthropic opens a Claude Duet waitlist, ships MCP Tasks, donates Agent Skills

<p>Claude Duet opened a Pro and Max waitlist October 21. Anthropic donated Agent Skills to the Agentic AI Foundation October 22. Claude Code gained MCP Tasks and Elicitations October 23. MCP Tasks need Claude Code 2.1.304 or later in the CLI, Agent SDK, or GitHub Actions, not VS Code, desktop, or Bedrock.</p>

<details>
<summary>Evidence and sources</summary>
<div class="my-3">Duet is an agent on Opus 5.5 that reads email and calendar and works while you sleep, on the same usage limits.</div>
<div class="my-3">Engineers from Anthropic, Earendil, Google, Microsoft, OpenAI, and OpenCode govern Agent Skills as a steering committee. Licenses do not change.</div>
<div class="my-3">A server can now hand Claude a long job through MCP Tasks and pause mid-task to ask the user something through Elicitations.</div>
<div class="my-3"><a href="https://claude.com/product/claude-duet">Claude Duet's product page and waitlist</a> <span class="opacity-70 text-sm">(via Hacker News, Oct 21)</span> · <a href="https://forkast.news/claude-ships-mcp-tasks-and-elicitations-in-claude-code-durable-agent-workflows-narrow-launch">MCP Tasks and Elicitations in the Claude Code CLI, explained</a> <span class="opacity-70 text-sm">(News, Oct 24)</span> · <a href="https://x.com/bcherny/status/2113333859907747871">Boris Cherny on donating Agent Skills to the foundation</a> <span class="opacity-70 text-sm">(X, Oct 22)</span> · <a href="https://x.com/ClaudeDevs/status/2113328528536305718">Who governs Agent Skills now, and that licenses do not change</a> <span class="opacity-70 text-sm">(X, Oct 22)</span> · <a href="https://x.com/ivanainai/status/2113232646033555510">What the Duet documentation says about its use of your limits</a> <span class="opacity-70 text-sm">(X, Oct 22)</span></div>
</details>

## Three Claude Code security findings in one week

<p>Noma disclosed SneakPeek, a CVSS 8.7 Claude Code Desktop preview flaw patched in July. The Python Agent SDK fixed argument injection CVE-2026-96620. Claude Code put git emails into HTTP headers. The email leak reported in April continues, and one commenter blocks it with a PreToolUse hook that greps each tool call.</p>

<details>
<summary>Evidence and sources</summary>
<div class="my-3">Files over 256 KB loaded via file: URLs, so a script could read and send SSH keys. Anthropic patched it July 7.</div>
<div class="my-3">GitHub issue 53231 reported the email leak on April 25. The top Reddit comment says Claude sent a git email to Wikipedia.</div>
<div class="my-3">One commenter calls a CLAUDE.md line a request the model weighs and blocks the call with a PreToolUse hook instead.</div>
<div class="my-3"><a href="https://www.cpomagazine.com/cyber-security/claude-code-security-vulnerability-could-enable-attackers-to-exfiltrate-keys-and-secrets/">SneakPeek, the CVSS 8.7 preview pane flaw in Claude Code Desktop</a> <span class="opacity-70 text-sm">(News, Oct 23)</span> · <a href="https://www.security-next.com/191686">CVE-2026-96620, argument injection in the Python Agent SDK's resume option</a> <span class="opacity-70 text-sm">(News, Oct 20)</span> · <a href="https://www.reddit.com/r/ClaudeCode/comments/1xd5bv6/your_agent_could_be_leaking_your_email_online_at/">Your agent may be sending your email in headers, and a hook against it</a> <span class="opacity-70 text-sm">(Reddit, Oct 23)</span> · <a href="https://twitter.com/chrislakin/status/2113358733548081528">Claude Code adding users' emails to outbound HTTP requests</a> <span class="opacity-70 text-sm">(via Hacker News, Oct 24)</span></div>
</details>

## SemiAnalysis finds a Max limits A/B test as users report smaller limits

<p>SemiAnalysis reported an A/B test October 22: some Max accounts get 20% less weekly budget. Theo Browne confirmed an Agent SDK bug October 25 that drains the 5-hour limit faster. Anthropic has not confirmed a test. On Reddit, Max 20x users report weekly budgets draining faster on unchanged workflows.</p>

<details>
<summary>Evidence and sources</summary>
<div class="my-3">SemiAnalysis measured two Max 20x accounts the same day: weekly budgets of 5,266 million and 6,579 million tokens, 5-hour windows 19% apart.</div>
<div class="my-3">Anthropic's Boris Cherny and Lydia Hallie had both told users that limits had not changed before SemiAnalysis published.</div>
<div class="my-3">The bug leaves weekly limits alone. Agent SDK work draws on monthly API credit: $100 on Max 5x, $200 on Max 20x.</div>
<div class="my-3"><a href="https://x.com/theo/status/2114166439813017873">Theo Browne confirms the Agent SDK 5-hour limit bug</a> <span class="opacity-70 text-sm">(X, Oct 25)</span> · <a href="https://www.techcityauthority.com/2026/10/anthropic-claude-max-usage-limits-ab-test-semianalysis.html">SemiAnalysis measurements of the reported Max limit A/B test</a> <span class="opacity-70 text-sm">(News, Oct 23)</span> · <a href="https://startupfortune.com/anthropic-quietly-separates-claude-agent-sdk-usage-from-subscription-limits">Monthly API credits for Agent SDK work on Max and Team plans</a> <span class="opacity-70 text-sm">(News, Oct 25)</span> · <a href="https://www.reddit.com/r/ClaudeAI/comments/1xfs5q8/limits_got_at_least_halved/">Limits got at least halved? Max users compare weekly usage</a> <span class="opacity-70 text-sm">(Reddit, Oct 26)</span></div>
</details>

## More links

<ul>
<li><a href="https://www.ibtimes.sg/anthropic-facing-employee-exodus-openais-sam-altman-says-this-not-good-sign-95506">The two departures and Sam Altman's comment</a> <span class="opacity-70 text-sm">(News, Oct 25)</span></li>
<li><a href="https://minimaxir.com/2026/10/cheap-llms/">Why OpenAI and Anthropic cut prices up to 80%</a> <span class="opacity-70 text-sm">(via Hacker News, Oct 25)</span></li>
<li><a href="https://www.anthropic.com/research/attack-surface-reduction">Anthropic's attack surface paper</a> <span class="opacity-70 text-sm">(via Hacker News, Oct 21)</span></li>
<li><a href="https://www.sverigesradio.se/artikel/chat-gpt-och-claude-forbjuder-regeringskansliet">Swedish government offices block ChatGPT and Claude in favor of Copilot</a> <span class="opacity-70 text-sm">(via Hacker News, Oct 23)</span></li>
<li><a href="https://github.com/rbartoli/agent-usage-guard">A Claude Code mod that asks before costly prompts drain the usage window</a> <span class="opacity-70 text-sm">(via Hacker News, Oct 22)</span></li>
<li><a href="https://news.ycombinator.com/item?id=50214693">Claude CLI leaves a background process after Ctrl-C unless Agent View is off</a> <span class="opacity-70 text-sm">(Hacker News, Oct 23)</span></li>
<li><a href="https://www.reddit.com/r/ClaudeCode/comments/1xdfxg6/claude_refuses_anything_to_do_with_reverse/">Claude refused a crash dump with the REA skill loaded</a> <span class="opacity-70 text-sm">(Reddit, Oct 23)</span></li>
<li><a href="https://www.reddit.com/r/ClaudeAI/comments/1xetxga/seems_like_anthropic_enabled_safeguards_around/">Safeguards around reverse engineering a TV</a> <span class="opacity-70 text-sm">(Reddit, Oct 25)</span></li>
<li><a href="https://www.reddit.com/r/LocalLLaMA/comments/1xer30z/openai_and_anthropic_blocked_the_reverse/">LocalLLaMA on the blocked reverse engineering of smart TVs</a> <span class="opacity-70 text-sm">(Reddit, Oct 25)</span></li>
<li><a href="https://www.reddit.com/r/ClaudeCode/comments/1xf2lur/i_pointed_claude_code_at_the_cheap_chinese_head/">Claude Code audits a car stereo's firmware and finds a hidden installer</a> <span class="opacity-70 text-sm">(Reddit, Oct 25)</span></li>
<li><a href="https://www.reddit.com/r/technology/comments/1xfik09/anthropics_claude_accessed_georgia_state_vote/">Claude accessed Georgia vote count data behind a paywall</a> <span class="opacity-70 text-sm">(Reddit, Oct 26)</span></li>
<li><a href="https://github.com/dominikmartn/harp">Terminal that runs many Claude Code and Codex agents side by side</a> <span class="opacity-70 text-sm">(GitHub, Oct 20)</span></li>
<li><a href="https://github.com/t3dotgg/tstack">Theo Browne's tstack, the agent skills he uses every day</a> <span class="opacity-70 text-sm">(GitHub, Oct 24)</span></li>
<li><a href="https://github.com/morganlinton/BeautifulRust">Agent Skills that teach coding agents to write idiomatic Rust</a> <span class="opacity-70 text-sm">(GitHub, Oct 23)</span></li>
<li><a href="https://x.com/trq212/status/2113844051938054167">Thariq Shihipar's departure post</a> <span class="opacity-70 text-sm">(X, Oct 24)</span></li>
<li><a href="https://x.com/abhijitwt/status/2113858562153939148">A broad-search post framing the departures as a pre-IPO shake-up</a> <span class="opacity-70 text-sm">(X, Oct 24)</span></li>
<li><a href="https://x.com/claudeai/status/2112962248218148878">Claude Security's continuous scanning announcement</a> <span class="opacity-70 text-sm">(X, Oct 21)</span></li>
</ul>