Gemini Memory, GPT-4o is Back | Weekly Edition

PLUS HOT AI Tools & Tutorials

Aug 15, 2025

Hey there! Welcome to your latest Creators’ AI Edition

This week, Google brings memory and temporary chats to Gemini, Anthropic’s Claude Sonnet 4 jumps to a 1M-token context window, and OpenAI rolls back changes by restoring GPT-4o and boosting GPT-5 limits. The talent wars heat up as Microsoft targets Meta researchers, Perplexity makes a bold bid for Google Chrome, and OpenAI once again takes gold, now at the IOI. Plus, there’s Musk vs. Apple drama, fresh AI tools from analytics to autonomous outreach, and step-by-step guides to master the latest models.

Let’s jump in!

Featured Materials 🎟️
News of the week 🌍
Useful tools ⚒️
Weekly Guides 📕
AI Meme of the Week 🤡
AI Tweet of the Week 🐦
(Bonus) Materials 🎁

Featured Materials 🎟️

Gemini Memory and Temporary Chats

Google has updated the Gemini app to provide a more personal and proactive assistant experience while giving users greater control over their data. A new feature allows Gemini to reference past chats to learn preferences and deliver tailored responses. This option is on by default, but can be turned off in settings. The company also introduced Temporary Chats, which do not affect personalization, are excluded from chat history, and are stored for up to 72 hours.

In addition, Google is renaming the Gemini Apps Activity setting to Keep Activity. When enabled, a sample of future uploads can be used to improve Google services. New controls also allow users to decide whether audio, video, and screen shares are used for service improvement. These updates are rolling out gradually, starting with the 2.5 Pro model in select countries, with more regions and models to follow.

New Claude Context Window

Anthropic’s Claude Sonnet 4 now supports up to 1 million tokens of context, allowing developers to process entire codebases of over 75,000 lines or dozens of research papers in a single request. The upgrade, now in public beta via the Anthropic API and Amazon Bedrock (with Google Cloud’s Vertex AI coming soon), unlocks new use cases in large-scale code analysis, document synthesis, and context-aware agents. Pricing adjusts for prompts over 200K tokens, with savings possible via prompt caching and batch processing. Early adopters like Bolt, new, and iGent AI report significant boosts in production-scale engineering and autonomous AI workflows.

News of the week 🌍

GPT-4o is Back

OpenAI CEO Sam Altman announced major adjustments to ChatGPT following user criticism of the GPT-5 launch. The GPT-4o model is returning to the model picker for all paid users, with assurances of advance notice before any future removal. Weekly limits for advanced reasoning in GPT-5 have increased from 200 to 3,000 queries, and the company clarified the model’s 196k context window.

New controls now let users select between Auto, Fast, and Thinking modes to ensure tasks are routed to the preferred model. Altman also revealed that a personality update for GPT-5 is on the way and stressed the importance of per-user customization and model personality. These changes aim to address concerns from a rollout that, despite strong technical progress, frustrated users with limited choice and routing issues.

OpenAI Achieved Gold Medal Again

OpenAI announced that its reasoning model achieved a gold-level score at the 2025 International Olympiad in Informatics (IOI), ranking sixth among humans and first among AI, despite having no programming-specific fine-tuning and relying only on basic tools. Scoring in the 98th percentile, up from 49% a year ago, the same model also won gold at the International Math Olympiad and AtCoder, underscoring rapid gains in AI reasoning that signal the approaching end of human dominance in such contests and hint at future breakthroughs in science, mathematics, and physics.

Share this post with friends, especially those interested in AI!

Elon Musk Against Apple

Elon Musk said on X that xAI is suing Apple for allegedly favoring OpenAI in the App Store and suppressing rivals like Grok, a claim quickly challenged with examples like DeepSeek and Perplexity reaching the top spot. The spat escalated when Sam Altman accused X of similar favoritism, prompting Musk to cite view count disparities and Altman to respond with “skill issue” and “or bots.” Grok chimed in supporting Altman and referencing Musk’s past algorithm tweaks, after which Musk shared a GPT-5 output ranking him more trustworthy than Altman while noting plans to reduce Grok’s dependence on legacy media.

Preplexity Want To Buy Google Chrome

Perplexity has reportedly offered $34.5B to buy Google Chrome, well below the browser’s estimated $50B value, after launching its own AI browser Comet. If accepted, Perplexity says it will invest $3B into Chrome and Chromium over two years, keep Chromium open source, and preserve users’ default settings. Chrome’s 3.5B users and 68% market share have drawn regulatory heat, with the US DOJ winning an antitrust case that could force Google to sell Chrome. While Google plans to appeal, rivals including OpenAI are reportedly eyeing bids.

Talent Wars Continue

Microsoft is reportedly targeting Meta’s AI talent with multi-million-dollar offers, matching compensation, and deploying “special recruiting teams” to lure researchers from units like Reality Labs, GenAI Infra, and Meta AI Research. The effort, led by Mustafa Suleyman and ex-Meta engineer Jay Parikh, includes a streamlined hiring process for “critical AI talent” with approvals in under 24 hours. While matching Mark Zuckerberg’s pay packages will be challenging, cultural issues within Meta’s AI unit may make some non-Superintelligence Labs staff more open to leaving.

Useful tools ⚒️

nFactorial AI- Video calls with the world's best minds as your personal tutors

mcp-use - Open source SDK and infra for MCP servers & agents

Sellinger AI - Autonomous AI LinkedIn Outreach

Macaron AI - The AI that instantly gets you and cooks up mini-apps

Airbook AI - Cursor for Analytics

Airbook AI connects to your CRMs, marketing, support, and product databases - understands your schema, auto-writes SQL, joins across sources, creates charts, builds dashboards, and activates workflows - In one place, on command.

Thanks for reading! This post is public so feel free to share it.