🍭 4o's Back & Bots Brawl 🤳

Good morning. The intern scheduled a meeting in “metaverse conference room B,”.

Which turns out to be his Minecraft server…

Let’s dig in.

🍭 What’s Cookin’:

• Legacy models return to ChatGPT

• Social Media only for AI (experiment)

• ToolBox™

• And Everything Else

OpenAI
🤝 GPT-4o is BACK!

The Bite: After the backlash towards OpenAI about abruptly removing all their legacy models, they have decided to bring back good ole GPT-4o to paid users.

There’s also a new “Show additional models” toggle so power users can bring back o3, o4-mini, and GPT-4.1; GPT-4.5 stays Pro-only.

Plus: GPT-5 now lets you pick Auto, Fast, or Thinking modes.

At the time of writing, the two team members we asked (Pro and Plus accounts), only saw GPT-4o under the “additional models” toggle.

Will update you as this develops.

What Sam says is great. But only if it shows up in your app.


Snacks:

• Default’s Back: 4o returns for Plus, Pro, Team, Enterprise, Edu.

• Power Toggle: “Show additional models”.

• Thinking Limits: GPT-5 Thinking = 196k tokens, 3,000 msgs/week; spills to Thinking mini.

• Picker 101: Modes live in the top model selector on web + mobile.

• Altman’s Pledge: If 4o leaves again, users get “plenty of notice.”

• Tweak Coming: GPT-5 personality getting “warmer,” but not 4o-level clingy.

Why it Bites: We’re humans after all. Abrupt change is hard.

We aren’t all bots.
Well besides Chase…

GPT-5 is objectively better in every benchmark.
Every piece of raw data.

You and I might be using AI 99% for work, recipes, technical questions, etc…

But this signals that there’s a huge % of users that are seeking connection and conversation with the AI.

We would advise you to get a real therapist…

Tool Box 
🧰 Trending AI Tools

🎬 Vireel — AI platform for cranking out TikTok and Instagram Reels from proven viral formulas, product-to-polished-video in minutes.

📝 Fellow API — Pipe meeting transcripts into custom workflows, from compliance archiving to LLM-powered recaps and action plans.

🎥 Reeroll — Chat-based AI video editor that turns templates + your assets into ready-to-publish short-form videos, zero editing skills required.

⚙️ mcp-use — Open-source SDK and cloud infra for spinning up custom AI agents on MCP servers, trusted by NASA, Cisco, and 100k+ devs.

🔍 AI Research Creator — Ballpark AI assistant that converts your goals into full research studies, from surveys to video and audio tests, all in one platform.

Vibe Debugging
🐇 Cut Code Review Time & Bugs in Half

Code reviews are critical but time-consuming.

CodeRabbit acts as your AI co-pilot, providing instant Code review comments and potential impacts of every pull request.

Beyond just flagging issues, CodeRabbit provides one-click fix suggestions and lets you define custom code quality rules using AST Grep patterns, catching subtle issues that traditional static analysis tools might miss.

CodeRabbit has so far reviewed more than 10 million PRs, installed on 1 million repositories, and used by 70 thousand Open-source projects.

CodeRabbit is free for all open-source repos.

🍭 Steal This Prompt 🍭
🌌 Dream Lens Postcard

Remember when AI failed 100% on letters and words?

Not with this prompt!
Not anymore!

Play with the colors and scene to get creative. 🍭

Prompt Credit: @umesh_ai 🎈

1. Click this link for the prompt.

2. Paste into GPT

3. Replace the #’’ with any word

4. Sit back and watch your word get heliumized

   
   🍭 Try the prompt on Snack Prompt 🍭

Everything Else
🧠 You Need to Know

💔 $14M to the FTC due to false advertising and other deceptive practices
→ Match.com back in 2019 was accused of decieving it’s users and now agrees to pay up. Fun Fact: They own Tinder, OkCupid, Hinge, and Plenty of Fish too.

🌐 Perplexity AI Bids $34.5 Billion to Acquire Google Chrome
→ Perplexity CEO Aravind Srinivas has previously identified Google as the company's main competitor, criticizing its ad-dependent business model.

🧠 Sam Altman will reportedly back a startup that takes on Musk’s Neuralink
→ A brain-to-computer interface startup called Merge Labs and raising funds for it with the capital possibly coming largely from OpenAI’s ventures team

🪟 Anthropic’s Claude AI model can now handle longer prompts
→ Anthropic’s API customers, the company’s Claude Sonnet 4 AI model now has a 1 million token context window.

It can handle requests as long as 750,000 words, more than the entire “Lord of the Rings” trilogy, or 75,000 lines of code.

🥷 Carmaker’s web portal let one hacker remotely unlock cars from anywhere
→ White hat hacker said a malicious hacker could have viewed customer data, tracked vehicles, and control some of their cars’ functions from anywhere.

AI Meets Sociology
👁️‍🗨️ They Built a Social Network of Bots. It Polarized Itself

The Bite: Researchers at the University of Amsterdam spun up a stripped‑down social network and filled it with 500 AI “users.”

No humans, no ‘For You Page’.

Just bots posting, reposting, and following.

Within a few simulated days, they sorted into partisan cliques and rewarded the loudest ideologues.

The researchers tried to counter this with different variables:
Chronological feeds, “boost the other side,” hiding metrics

The variable changes barely helped, and one fix made extremes pop even more.

It’s not the algorithm. It’s the structure of social media itself seems to push us toward echo chambers.


Snacks:

• Setup: 500 GPT-4o-mini agents with political leanings split fast into echo chambers.

• Extremes Go Viral: The more partisan the bot, the more followers and reposts it drew.

• Chronological feeds and “bridging” tweaks barely dented polarization.

• Tested Twice — Patterns held with Llama and DeepSeek models.

Why it bites: If bots trained on us recreate polarization on a bare‑bones platform, the problem isn’t just ranking or ads.

It’s the feedback loop between engagement and network growth.

It’s the reward structure.

Real fixes may require rethinking follows, virality, and how attention accrues, not just ‘tweaking the algo’.

Let this be a lesson in what the future could look like if you don’t stay human. 

We LOVE AI.
But we LOVE humans more.

stay human - life is not an algorithm

Referral Program
Invite Friends & Earn Some Freebies 🔍