Good morning!

Anthropic warns that AI systems may soon help build better versions of themselves faster than humans can control, while OpenAI launched a new β€œdreaming” memory system that makes ChatGPT more personal over time.

In todays email:

  • Daily Update

  • Social Media

  • Today’s Highlight

  • YouTube

  • Today Trend

  • Prompt

Read time: 7 min

DAILY UPDATE

80% of New Code Now Written by AI

Anthropic says AI may soon improve itself faster than humans can control it.

Anthropic released a new report called β€œWhen AI builds itself,” warning that recursive self improving AI could arrive sooner than expected. The company said its AI model Claude is already helping speed up AI development at a major level.

  • More than 80% of Anthropic’s merged code in May was written by Claude, with engineers now producing 8x more code per day compared to 2024.

  • Anthropic researcher Jack Clark said future versions of Claude may eventually build the next version completely without human help.

  • Both Anthropic and OpenAI are now discussing possible slowdowns and policy talks as concerns grow around fast moving AI progress.

This matters because more AI labs are seeing early signs of self improving AI systems. Companies like MiniMax also say their models helped build themselves, while new startups are focusing only on this idea.
Continue Reading…

TODAY’s HIGHLIGHT

ChatGPT Gets β€œDreaming” Memory That Learns You Over Time

OpenAI adds new β€œdreaming” memory system that makes ChatGPT more personal and smarter over time

OpenAI has launched a new memory upgrade called β€œdreaming.” It helps ChatGPT build a clear, ongoing profile of each user. This makes chats more personal and better at remembering context across time. Instead of saving random facts, ChatGPT now keeps a structured summary of users. It groups details into areas like work, travel, and hobbies.

  • ChatGPT now builds a running profile grouped by topics like work and hobbies

  • Users can edit, correct, or delete stored memories anytime

  • Tests show big gains: accuracy jumped from 41.5% to 82.8%, and preference matching rose to 71.3%

  • The update is rolling out first to Plus and Pro users in the US, with more regions coming soon

This matters because memory makes AI feel more useful and personal. It helps ChatGPT give answers that fit the user better over time. It also makes switching to another AI harder, since the system slowly becomes more tied to each person’s history and needs.
Continue Reading…

YOUTUBE

AI Scams on Amazon

TODAY TREND

Leni
The world’s most accurate AI for investors

SellerClaw
A team of AI agents that runs your stores across channels

Minimi
Your ambient memory for Claude

Veltrix AI
AI finance copilot for cash flow, margins, and growth

Ideogram 4.0
Generate design-ready image with open weight, layout control

SOCIAL MEDIA

That’s it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve the experience for you.

You can unsubscribe from here

Login or Subscribe to participate

Keep Reading