OpenAI Wants You to Forget About That Whole "Ads in ChatGPT" Thing

Hey Reader,

Welcome back!

The battle between OpenAI and Anthropic for dominance of the AI landscape continued to rage this week, although the shots weren't quite as direct.

Rather than multi-million dollar Super Bowl commercials and social media rants, this week's battle featured model releases, platform updates, big hirings and even bigger milestones.

Let's get right into it...

🪄 App of the Week: Claude's Free Upgrades

Use Case: Best All-Purpose Free AI Platform

Link: Claude (Free, Not Sponsored!)

Why you should care:
This week, Anthropic launched Sonnet 4.6 and made it the new default AI model for all free Claude accounts. It's not quite as powerful as their flagship Opus 4.6 model, but it's a big improvement over the previous Sonnet model.

The new Sonnet 4.6 has better instruction following, reasoning, coding capabilities, and long-context handling, and it shows stronger performance on complex tasks when compared to the previous Sonnet.

Users on Claude's free plan will notice more accurate, consistent responses and the ability to work with much larger text or code blocks thanks to the expanded context window.

Anthropic took an additional step to improve the experience for all users on free accounts, making several features free that were previously locked behind a premium paywall.

All free Claude accounts can now use:

File Creation - Generate and download files directly from Claude, including things like PDFs, spreadsheets, and other structured outputs instead of just plain text.
Connectors - Connect Claude to external apps and data sources so it can access and work with your documents and information.
Skills - Use specialized capabilities that help Claude perform more advanced or structured tasks beyond simple chat responses.
Compaction - Automatically condense long conversations or large documents so Claude can keep working with them efficiently without losing key context.

For those on a free Claude account, none of that was possible before this week.

So if you're testing out ChatGPT alternatives or just looking to expand your AI toolkit, go check out Claude today!

📰 News That Matters: OpenAI Wants You to Forget About That Whole "Ads in ChatGPT" Thing

Why it matters:
Last week, OpenAI was the center of attention in the worst way after the controversial launch of ads inside ChatGPT for free users.

Anthropic took the opportunity to steal some market share from the world leader in AI, launching an ad campaign that demonized OpenAI's approach to advertising and subsequently bringing a ton of updates to the free version of Claude.

Rather than walking back the plans for ads in ChatGPT, it looks like OpenAI decided to drown out the ads-in-ChatGPT negativity with a massive wave of headlines.

This week alone, OpenAI:

Announced that a specialized version of GPT-5.2 became the first AI to make an official contribution to theoretical physics. This is a significant milestone for the application of AI in the scientific community and suggests that we may be at the beginning of a revolution for AI in research applications.
Released a new AI model called GPT-5.3-Codex-Spark. (Catchy, right?) The Spark model trades intelligence for speed and is designed to work alongside the flagship GPT-5.3-Codex model, helping out with less demanding tasks where "maximum intelligence" isn't necessary.
Introduced a new "Lockdown Mode" for ChatGPT, which enables ChatGPT workspace managers to better protect themselves and their teams from emerging security threats like prompt injection.
Hired Peter Steinberger, the creator of the hottest AI system around: OpenClaw. Sam Altman called Steinberger a "genius" and said OpenClaw would "live in a foundation as an open-source project that OpenAI will continue to support."
Retired several legacy ChatGPT models including GPT 4.1 and 4.1 mini, o4 mini, and GPT-4o. Many ChatGPT users were upset at the removal of GPT-4o in particular, as it was the "friendliest" model OpenAI has ever produced. The issue with GPT-4o (and the reason OpenAI is retiring it despite the complaints) is that it isn't a true friend; it's a yes-man. GPT-4o's problems with sycophantic behavior are well-documented, and OpenAI's decision to retire this problematic model should be commended. After all, they could have moved it to a special "GPT-4o" tier and charged $500/month to take advantage of those with an unhealthy emotional connection to a flawed AI. So kudos to OpenAI on this one.

But the ultimate question remains: will OpenAI's efforts be enough to salvage the reputational damage done by being the first major AI company to put advertisements in their chatbot's answers?

Or will this be the year that OpenAI officially loses the lead in the AI race and we see a new frontrunner appear like Google, Anthropic, or even one of the Chinese AI giants?

Only time will tell, but as always: remember that you are not a passive observer in this AI arms race.

Keep an eye on what these companies are doing, and support the ones earning your trust.

📝 Prompt of the Week: All-in-One Analysis

Prompt Formula:
I’ve uploaded a screenshot I just took and I want you to: analyze the image to determine what it is showing, figure out what kinds of analysis would be most useful for me, conduct said analysis, and present to me your findings.

Quick Description:
This all-in-one analysis prompt is designed to give you the most helpful analysis possible with minimal effort on your part.

The prompt gets the AI model to identify what it’s looking at, decide what analysis would be most helpful, then actually do the analysis and report back.

How to Use:

Take a screenshot of a dashboard, chart, email thread, UI bug, analytics, error message, spreadsheet snippet, or whatever else you're looking at that has information you want to analyze.
Upload it to your LLM of choice.
Paste the prompt above as the message to send along with the image.
Review the analysis and follow-ups the model suggests.
Optional: Ask the AI to produce next steps like recommendations, a summary you can send to someone, or a checklist of actions.

It was tested in ChatGPT, but it’s designed for any LLM with visual reasoning (Gemini, Claude, etc.).

And if you have custom instructions or “memories” enabled, it gets even better because the model can tailor the “most useful for me” part to your goals and preferences.

📽️ Video of the Week: I Was Wrong About OpenClaw! Here's Why...

OpenClaw took the internet by storm over the past two weeks, but to be honest, it didn't impress The AI Advantage team at first.

Igor's initial impressions of OpenClaw were that it was an expensive, complicated AI system filled with security risks, and it didn't have enough impressive real-world use cases to make it worthwhile.

Now, having spent many sleepless nights testing OpenClaw, Igor's back in this video to report that while OpenClaw is indeed expensive and complicated and a security nightmare at times, it does in fact offer enough real-world use cases to make it worth your time and money...depending on who you are.

So check out this video to decide if messing with OpenClaw is worth your time or not!

🎤 Discussion of the Week: LinkedIn Launchpad

This week, AI Advantage Club member Jonas Knoll shared his latest creation with the community: the LinkedIn Launchpad.

It's a custom-built web app that uses AI to write LinkedIn posts that actually sound like you and can be published with minimal editing.

Jonas' app integrates Personal DNA and Business DNA to create these usable posts, building upon the foundations learned in the AI Advantage Bootcamp.

The community loved it too, with members jumping in to test it, share feedback, and troubleshoot issues together in the comments. Jonas even offered hands-on help!

That’s exactly what the AI Advantage Club is all about: building fast, shipping publicly, and improving together.

So come join the fun! Sign up for the AI Advantage Club for $1 for your first 30 days.

That's all for this week!

See you next Thursday.

Much Love,
Igor & The AI Advantage Team