AI & robotics news brief ·
Google loses $500 billion in a day
Good morning. Google lost $500 billion in one day, making it the clearest market-moving story in today’s headlines. There’s also a security warning around a malicious Claude Code extension, plus more ads are coming to Google’s AI Overviews.
Today in AI
Google Loses $500 Billion in One Day
Why it matters: Alphabet's significant loss of market value highlights the financial risks associated with heavy investments in AI technology, raising concerns about the sustainability of its business model.
Cursor Discloses SSH Backdoor Attempt in Malicious Claude Code Extension
Why it matters: The discovery of a backdoor in a widely used code extension highlights significant security vulnerabilities in software distribution channels, impacting user safety and trust.
Here’s Why You’re About to See More Ads in Google’s AI Overviews
Why it matters: The expansion of ads in Google's AI Overviews may reduce organic traffic to online publishers, challenging their business models and impacting how content is accessed online.
You don’t need to believe in a god to get a religious exemption from using AI at work
Why it matters: This story highlights a growing trend where individuals seek legal religious exemptions from the use of AI in the workplace, raising questions about technology's role and impact on personal beliefs.
Prime Intellect Releases Prime Mega MoE: Trainable Fused MoE Kernel, 3.11x...
Why it matters: The introduction of Prime Mega MoE enhances the efficiency of training mixture-of-experts models, potentially reducing the computational resources required for AI model development.
Robotics
Flexible Automation vs Agentic Augmentation in Coding
Why it matters: The evolution of coding practices through agentic automation and flexible automation could significantly impact developer productivity and skill development, shaping the future of the software industry.
Trump declares robot imports a national security threat
Why it matters: This story highlights a significant shift in U.S. trade policy which could affect the competitiveness and operational cost of American manufacturing due to potential tariffs on foreign-made robots.
Agent uses to try
Research & Learning · Codex
Use Codex to summarize Logseq journals from the CLI
They used Codex with the Logseq CLI docs to query a note graph and turn months of journals and tagged pages into summaries. The setup let Codex produce summaries from long stretches of journals, and tagged blocks made the searches feel straightforward. Wosk described the whole workflow as very alpha, and shared no counts or measurements.
Try this prompt
I’m using [agent name] with the Logseq CLI docs. Read the docs at [URL or path], connect to my Logseq graph at [path], and summarize journal entries from [date range] plus pages tagged [tag]. Include main themes, open tasks, and the current status of each project or application. Draft the summary in [file name], then pause and ask for my approval before you send, submit, delete, or change anything in my notes.
Money & Bills · Claude Cowork
Build weekly expense and tax summaries from your records
Have Cowork turn receipts, transactions, and email history into reports and maintenance logs. The tax summary came back in minutes rather than days, and a LinkedIn user said they saved about 5 hours. The home-maintenance log included dates, contractors, and costs. The source does not give accuracy figures for the finance tasks.
Try this prompt
Watch for receipts or transaction records I share at [email/folder/source]. Each [week/Monday], gather them into an expense summary with date, vendor, amount, and category, then stop and ask for my approval before you send or share anything. Also review [bank export / credit-card export] from [time period] and draft a tax-prep summary that separates income, deductible expenses, and items that need my review. Finally, scan [email folder / repair records] for home-maintenance history and make a schedule with dates, contractors, and costs, then ask before adding anything to my calendar.
Sales & Marketing · Devin
Triage client Slack messages into tickets and drafts
Use Devin to inspect new client messages, open Linear tickets, and draft replies for human approval. The workflow now handles customer-service triage, ticket creation, and reply drafting while keeping all client communication under human review. No message is sent on its own yet because the team is still tightening the wording.
Try this prompt
Watch the Slack channel [channel name] for new client messages. For each new message, inspect the issue using [repo, docs, or other sources], create a Linear ticket in [Linear project] with your findings, and draft a reply for the client. Post the ticket link and reply draft in [internal channel], then pause and wait for my approval before sending, submitting, deleting, or making any other irreversible change. After a ticket is fixed, route any follow-up messages through the same review step. Treat the task as complete only when the ticket exists and the draft is ready for review.