Good morning,

It was a loud week in AI. Anthropic raised at a $900 billion valuation, OpenAI quietly crossed $25B in ARR, Meta announced 8,000 cuts coming May 20, and on Friday the Freshworks CEO did something most CEOs avoid: he publicly attributed his 500-person layoff to AI. My honest take this morning is that everyone's about to focus on the layoffs, but the quieter story this week is that the model under your ChatGPT tab actually changed underneath you on May 5. That's the one I'd handle first. Story two is the one I'd think about all week.

— Khadin

In today's issue:

Hit reply and tell me what you think. I read every one.

The Scoop: OpenAI flipped GPT-5.5 Instant to default on May 5. The release notes claim 52.5% fewer hallucinated claims than GPT-5.3 on high-stakes prompts. Real-world reports are mixed, and any workflow tuned on the previous default is going to behave a little differently this week.

If your ChatGPT has felt off this week and you couldn't quite put a finger on why, the model under you probably moved. It's worth twenty minutes to check before you start blaming your prompt.

Open ChatGPT and Claude.ai side by side. Paste the same prompt you use most often into both. Look at which one feels right on your kind of work, and if there's a clear winner, switch your main tab. This isn't really an OpenAI versus Anthropic argument. It's that defaults are a quiet battleground in 2026 and most people are still running whichever model loaded first.

Anthropic has been pushing hard on the financial-services and long-context side this quarter. Claude Opus 4.7 dropped May 5 with new agent templates for things like pitchbook drafting and KYC screening. Project Glasswing reportedly gave AWS, Apple, Google, JPMorgan, and Microsoft early access to an unreleased "Mythos" model so they can find software bugs before code ships. The bigger pattern is that the frontier is moving on both sides, and the model that worked best for your workflow back in March might not be the one that works best in May.

So what for you: If your AI feels off this week, test before you blame yourself.

If you read about AI all week but you still haven’t figured out, how to actually use AI to automate tasks or achieve something.

You can describe your role and what you do to AI Insiders Assistant and it will help you automate upto 90% of your tasks w/ ai.

The Scoop: Freshworks announced a 500-person layoff on Friday, May 8. CEO Girish Mathrubootham was quoted in coverage saying over half the company's code is now AI-written. Meta's planned May 20 cuts are estimated at 8,000 people, roughly 10% of the company.

This one is worth marking down for a non-obvious reason. The Freshworks layoff isn't the biggest one this year. The thing that's different is that the CEO publicly attributed it to AI. Until now most layoff letters have buried that part. After Friday it's much harder to keep that buried.

Here's how I'd read it. If your job involves code or anything close to it, the next 90 days will probably include a 1:1 where someone asks how you're using AI. It's a question worth being ready for instead of one to dread.

The lowest-effort prep is a five-line note. Write down the AI tools you actually use day to day, the rough time you save on each, and add one example you could put on a screen if asked. Five minutes to write. The note shows up well in a review, in a layoff conversation if it ever happens, and frankly even if you're just asking for a raise this quarter.

So what for you: Open a doc this morning, write the five lines, save it. Done before your second coffee.

→ Meta's May 20 cuts. If the layoff letter names AI the way Freshworks did, the new norm is set. Watch the wording on Wednesday or Thursday once the press leaks start.

→ Microsoft Agent 365 ($15 per user per month, launched May 1). Most companies haven't budgeted for this yet. The CTOs who ship internal memos this week are the ones whose teams will be retooling first.

→ Anthropic's Project Glasswing. AWS, Apple, Google, JPMorgan, and Microsoft reportedly got early access to "Claude Mythos Preview" to find software bugs before code ships. Two reads on this. One, where frontier security review is heading. Two, which companies Anthropic actually trusts.

The Scoop: Lindy lets you build AI assistants that read your inbox, book meetings, qualify leads, and run support across 4,000+ apps. No code required. Default model is Claude Sonnet 4.5, the same one that recently hit 77.2% on SWE-bench coding tasks.

What makes Lindy different from a generic AI assistant:

→ Use Lindy.ai free for 7 days. Full access to inbox management, meeting scheduling, and the Gaia voice agent. No credit card to start.