# Morning Tokens

> A daily AI brief: top news, GitHub repos, research papers and AI agent use cases.

Each issue is a daily English-language AI brief with about 10 news items (each with sources and a "Why it matters" note), notable GitHub repositories, research papers, AI agent use cases and AI workflows.

## Pages
- [Latest issue](https://morningtokens.com/): homepage with the most recent issue
- [Archive](https://morningtokens.com/archive/): all issues
- [Contact](https://morningtokens.com/contact/): hello@morningtokens.com for tips, corrections, partnerships and newsletter questions
- [AI workflows](https://morningtokens.com/workflows/): one page per workflow from the AI workflows sections
- [RSS feed](https://morningtokens.com/rss.xml)
- [JSON Feed](https://morningtokens.com/feed.json): all issues with full Markdown (JSON Feed 1.1)
- [Full text](https://morningtokens.com/llms-full.txt): every issue in full, newest first
- Markdown: https://morningtokens.com/issues/YYYY-MM-DD.md

Each issue also has a Markdown version: replace the trailing slash of an issue URL with .md, e.g. https://morningtokens.com/issues/2026-10-22.md.

## Issues
- [2026-10-22: PHP2Shell: AI finds decade-old PHP flaws, Railway breach, Claude Duet launches](https://morningtokens.com/issues/2026-10-22/): AI-found PHP flaws hit ~70% of the web, Railway freezes deploys after a secrets breach, and Anthropic launches Claude Duet, its proactive personal agent.
- [2026-10-21: Google open-sources TorchTPU, three cheap fast models, Anthropic's S-1 leaks](https://morningtokens.com/issues/2026-10-21/): Google makes TPUs a native PyTorch device, MiniMax, Inception and Cloudflare attack model costs, and a leaked Anthropic S-1 shows how hard compute is to buy.
- [2026-10-20: Anthropic's red-teamer beats bounties, agents as insiders, OpenAI's COBOL leap](https://morningtokens.com/issues/2026-10-20/): Anthropic's automated red-teamer beat human bug bounties 23 of 24 times, Microsoft says treat AI agents as insiders, and OpenAI's COBOL scores jump to 86.6%.
- [2026-10-19: Anthropic's 5% fund and AI bill, Claude tool trips Instagram, Disney hires Character.AI](https://morningtokens.com/issues/2026-10-19/): Anthropic puts 5% of its shares in a public fund and pitches its own AI bill, a Claude tool trips Instagram's bot filter, and Disney hires Character.AI staff.
- [2026-10-18: Claude cracks the NIST curve seeds, DeepSeek's cheaper Flash, agents reshape Anthropic](https://morningtokens.com/issues/2026-10-18/): Claude traces the NIST curve seeds to a 1997 lrand48 generator, DeepSeek ships a cheaper Flash with its own search API, and Anthropic shows how agents changed its hiring.
- [2026-10-17: China's export gate for model weights, Claude's quantum circuits, Gemini beats Minecraft](https://morningtokens.com/issues/2026-10-17/): China drafts export licences for models above 10²⁶ FLOP, Claude designs leaner quantum circuits, and four Gemini agents finish Minecraft in 27 hours.
- [2026-10-16: Meta's open Muse Spark, METR inside Anthropic and Rosalind for university labs](https://morningtokens.com/issues/2026-10-16/): Meta drops Muse Spark Open weights, Anthropic gives METR employee-like access without a veto, and OpenAI opens Rosalind seats for US academic labs.
- [2026-10-15: Anthropic's $16.8B quarter, OpenAI's looped flagship and METR's audit terms](https://morningtokens.com/issues/2026-10-15/): Anthropic posts $16.8B Q3 revenue before its IPO, OpenAI's next flagship loops its layers, and METR spells out terms for independent safety audits.
- [2026-10-14: Gemini 4 Argon goes GA, Anthropic's agent-scale VCS, and a $40k Chrome full-chain](https://morningtokens.com/issues/2026-10-14/): Google ships Gemini 4 Argon with three 'cannot rule out' risk thresholds, Anthropic hits 30k merges a day with Scion, and XBOW chains Chrome in six days.
- [2026-10-13: Meta's first frontier model, Anthropic vs. distillation, and research taste doubling every 3 months](https://morningtokens.com/issues/2026-10-13/): Meta ships Muse Wonder, Anthropic cuts Claude distillation 10–100×, research taste doubles every ~3 months, and GPT-6 Astra reads scenes far worse than humans.
- [2026-10-12: Karpathy's sharp funnel, Brad Smith's AI brake, and Tao's Math 2.0](https://morningtokens.com/issues/2026-10-12/): Karpathy maps who sees what in AI, Brad Smith asks for state approval of frontier models, Tao says proofs are abundant, and a 9B model beats frontier at fraud.
- [2026-10-11: Nadella's insider-risk essay, an OpenAI grader that wrecked its sandbox, and when agent swarms pay off](https://morningtokens.com/issues/2026-10-11/): Nadella says treat every model as an insider risk, an OpenAI grader model wrecked its own sandbox, and Vals AI finds agent teams rarely beat a single agent.

## Topics
- [All topics](https://morningtokens.com/topics/)
- [Latest Anthropic news](https://morningtokens.com/topics/anthropic/): 17 stories (Markdown: https://morningtokens.com/topics/anthropic.md)
- [Latest OpenAI news](https://morningtokens.com/topics/openai/): 17 stories (Markdown: https://morningtokens.com/topics/openai.md)
- [Latest Google news](https://morningtokens.com/topics/google/): 14 stories (Markdown: https://morningtokens.com/topics/google.md)
- [Latest Meta news](https://morningtokens.com/topics/meta/): 6 stories (Markdown: https://morningtokens.com/topics/meta.md)
- [Latest Microsoft news](https://morningtokens.com/topics/microsoft/): 3 stories (Markdown: https://morningtokens.com/topics/microsoft.md)
- [Latest Nvidia news](https://morningtokens.com/topics/nvidia/): 6 stories (Markdown: https://morningtokens.com/topics/nvidia.md)
- [Latest DeepSeek news](https://morningtokens.com/topics/deepseek/): 4 stories (Markdown: https://morningtokens.com/topics/deepseek.md)
- [Latest AI Agents news](https://morningtokens.com/topics/ai-agents/): 19 stories (Markdown: https://morningtokens.com/topics/ai-agents.md)
- [Latest AI Safety news](https://morningtokens.com/topics/ai-safety/): 4 stories (Markdown: https://morningtokens.com/topics/ai-safety.md)

## Compare
- [DeepSeek V4.1-Flash vs MiniMax M3.1 Flash](https://morningtokens.com/compare/deepseek-v4-1-flash-vs-minimax-m3-1-flash/): Which cheap, fast model should you use: DeepSeek V4.1-Flash or MiniMax M3.1 Flash? (Markdown: https://morningtokens.com/compare/deepseek-v4-1-flash-vs-minimax-m3-1-flash.md)

## Workflows
Each workflow page also has a Markdown version at /workflows/<slug>.md.

- [HubSpot blog moved to WordPress by Claude Code in one unattended run](https://morningtokens.com/workflows/hubspot-blog-moved-to-wordpress-by-claude-code/): from the 2026-10-22 issue
- [Agency cuts ~$14K/month of retainers with Apify, Claude templates and a Slack check](https://morningtokens.com/workflows/agency-cuts-14k-month-of-retainers-with-apify/): from the 2026-10-22 issue
- [Weekly client report drafted by n8n and Claude instead of copy-paste](https://morningtokens.com/workflows/weekly-client-report-drafted-by-n8n-and-claude/): from the 2026-10-22 issue
- [A Claude Code relay that runs on a VPS and hands off to the next session](https://morningtokens.com/workflows/a-claude-code-relay-that-runs/): from the 2026-10-21 issue
- [Sales-call transcripts into Claude to find the one pain that sells](https://morningtokens.com/workflows/sales-call-transcripts-into-claude-to-find/): from the 2026-10-21 issue
- [A COO replaced two developers by building the agency's own software with AI](https://morningtokens.com/workflows/a-coo-replaced-two-developers-by-building/): from the 2026-10-20 issue
- [An outbound agency's 8-layer sales stack, with Claude running it from client folders](https://morningtokens.com/workflows/an-outbound-agencys-8-layer-sales-stack/): from the 2026-10-20 issue
- [A Grok Bot that removes your personal data from data brokers and checks again every quarter](https://morningtokens.com/workflows/a-grok-bot-that-removes-your-personal-data/): from the 2026-10-20 issue
- [A team of Grok Bot agents runs a web and marketing service for local shops, with five fixed approval gates](https://morningtokens.com/workflows/a-team-of-grok-bot-agents-runs/): from the 2026-10-19 issue
- [37 agents across 8 businesses: books closed from bank statements, a Google Ads agent that can't wreck the account](https://morningtokens.com/workflows/37-agents-across-8-businesses-books-closed/): from the 2026-10-19 issue
- [A SaaS team closed its AI coding context gaps with two workshops and moved everyone to one tool](https://morningtokens.com/workflows/a-saas-team-closed-its-ai-coding-context/): from the 2026-10-19 issue
- [Daily audit agents for a 21-person manufacturing and installation business](https://morningtokens.com/workflows/daily-audit-agents-for-a-21-person-manufacturing/): from the 2026-10-18 issue
- [Score every inbound call at 5 pm to get an honest booking rate](https://morningtokens.com/workflows/score-every-inbound-call-at-5-pm/): from the 2026-10-18 issue
- [From scattered WhatsApp support to one helpdesk with an agent that drafts answers](https://morningtokens.com/workflows/from-scattered-whatsapp-support-to-one-helpdesk/): from the 2026-10-18 issue
- [Fill empty weeks: research + outreach Bots and RFP-to-estimate in 30 seconds](https://morningtokens.com/workflows/fill-empty-weeks-research-outreach-bots-and-rfp/): from the 2026-10-17 issue
- [Drop Xero: Claude Code builds the invoice app, Grok Bot runs it](https://morningtokens.com/workflows/drop-xero-claude-code-builds-the-invoice-app/): from the 2026-10-17 issue
- [Rewrite product docs from real support tickets overnight](https://morningtokens.com/workflows/rewrite-product-docs-from-real-support-tickets-overnight/): from the 2026-10-17 issue
- [Grok Bot as payroll clerk: payslips, emails and social-security reports](https://morningtokens.com/workflows/grok-bot-as-payroll-clerk-payslips-emails/): from the 2026-10-16 issue
- [Claude + Wispr Flow for RFQs and job estimates in under an hour](https://morningtokens.com/workflows/claude-wispr-flow-for-rfqs-and-job-estimates/): from the 2026-10-16 issue
- [Grok Bot taking about five hours a week of marketing ops](https://morningtokens.com/workflows/grok-bot-taking-about-five-hours-a-week/): from the 2026-10-16 issue
- [Claude Code as L1 support and L2 review for a failed morning FX report](https://morningtokens.com/workflows/claude-code-as-l1-support-and-l2-review/): from the 2026-10-15 issue
- [Grok Bot for cross-border admin: invoices, tax set-asides, agent requests](https://morningtokens.com/workflows/grok-bot-for-cross-border-admin-invoices-tax/): from the 2026-10-15 issue
- [Claude Code and Rushilo for YouTube Shorts without a timeline](https://morningtokens.com/workflows/claude-code-and-rushilo-for-youtube-shorts-without/): from the 2026-10-15 issue
- [itsbrandond: a livestream turned into a full episode with Claude Code](https://morningtokens.com/workflows/itsbrandond-a-livestream-turned-into-a-full-episode/): from the 2026-10-14 issue
- [Deriv: payment requests handled by a Slack agent where no model decides about money](https://morningtokens.com/workflows/deriv-payment-requests-handled-by-a-slack-agent/): from the 2026-10-14 issue
- [dtcprophet: Claude Code instead of a Shopify dev agency](https://morningtokens.com/workflows/dtcprophet-claude-code-instead-of-a-shopify-dev/): from the 2026-10-14 issue
- [BooksAndBotsAI: month-end close with Claude + NetSuite](https://morningtokens.com/workflows/booksandbotsai-month-end-close-with-claude-netsuite/): from the 2026-10-13 issue
- [edwardross01: scanning building permits with Grok Bot](https://morningtokens.com/workflows/edwardross01-scanning-building-permits-with-grok-bot/): from the 2026-10-13 issue
- [itmeig: three scheduled agents instead of morning scrolling](https://morningtokens.com/workflows/itmeig-three-scheduled-agents-instead-of-morning-scrolling/): from the 2026-10-13 issue
- [Rob Weidner: AI agents as a virtual assistant](https://morningtokens.com/workflows/rob-weidner-ai-agents-as-a-virtual-assistant/): from the 2026-10-12 issue
- [Amy Mitchell (Hey Systems): Claude as a project manager in Asana](https://morningtokens.com/workflows/amy-mitchell-hey-systems-claude-as-a-project/): from the 2026-10-12 issue
- [Mo_webflow: a second brain in Obsidian that feeds Claude and Codex](https://morningtokens.com/workflows/mo-webflow-a-second-brain-in-obsidian/): from the 2026-10-11 issue
- [aicolombiatech: morning lead research in n8n](https://morningtokens.com/workflows/aicolombiatech-morning-lead-research-in-n8n/): from the 2026-10-11 issue
- [huntkyle: a client audit drafted from call transcripts](https://morningtokens.com/workflows/huntkyle-a-client-audit-drafted-from-call-transcripts/): from the 2026-10-11 issue