🍭 GPT-5.3: Instant Upgrade 🔥

Good morning. If AI replaces every job except “guy who sends slightly unhinged newsletters,” I will consider that a massive technological success.

Let’s dive in 👇

🍭 What’s Cookin’:

• OpenAI ships GPT-5.3 Instant, calls the last one “cringe”

• QuitGPT organizes backlash against OpenAI

• Brazil forces Meta to allow rival AI bots on WhatsApp

• NVIDIA drops $4B on photonics for next-gen AI infrastructure

QuitGPT
🪧 The Organized Backlash Against OpenAI

The Bite:

In the days following OpenAI's Pentagon deal, a structured boycott campaign called QuitGPT emerged as the organized face of user backlash.

The movement claims over 1.5 million people have taken action:

• cancelling subscriptions,

• sharing boycott content,

• or signing up to the campaign.

It recommends switching to Claude, Gemini, or open-source alternatives, and planned an in-person protest outside OpenAI's San Francisco HQ on March 3.

Snacks:

• QuitGPT describes itself as a response to OpenAI accepting terms Anthropic publicly refused on safety grounds

• The campaign explicitly warns against Grok as an alternative

• It argues most users don't know other AI assistants exist

• An in-person protest at OpenAI HQ was organized for March 3

• ChatGPT has ~900 million weekly users; 1.5 million "actions" are under 0.2% of that base

Why it Bites:

QuitGPT matters less as a boycott and more as a signal that AI companies now have organized civil opposition with a brand, a website, and a street presence.

That's new.

The actual defection numbers are tiny relative to ChatGPT's user base.

But the infrastructure being built matters: a campaign that redirects users, names alternatives, and coordinates protests.

The real audience for QuitGPT isn't current ChatGPT users.

It's regulators, journalists, and enterprise buyers watching to see whether AI safety concerns can generate sustained public pressure.

That test is still running.

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🧬 Aura
Semantic version control for AI code. Hashes the AST (not lines) to trace AI-generated changes, block undocumented commits, and rewind broken functions.

📊 ChatWithAds
Ask questions about your ads + business data and get actionable answers without drowning in dashboards and spreadsheets.

🧾 Didit
Modular identity verification with free unlimited KYC, plus add-ons like biometrics/AML/age checks when you need them.

🔎 Rankfender
Tracks and improves your brand’s AI answer visibility, generates keyword ideas, and can auto-publish/optimize SEO content (WP/Shopify/Wix).

🧱 CtrlAI
A transparent HTTP proxy that sits between your agent and the LLM to enforce guardrails, audit behavior, and block unsafe tool calls with zero SDK changes.

Brazil vs. Meta
⚔️ Brazil Just Forced Meta’s AI Market Open on WhatsApp

The Bite:

Brazil's antitrust tribunal CADE unanimously upheld a preventive measure blocking Meta from enforcing new WhatsApp Business terms of use that would have banned competing AI chatbots from the platform.

CADE’s first AI competition precedent forces Meta to restore access for competing AI providers like Luzia and Zapia within five days of official publication.

The investigation into whether Meta abused its dominant position in messaging to favor its own Meta AI continues.

Snacks:

• Meta's new WhatsApp Business terms, announced October 2025, would have banned third-party AI providers from the platform as of January 15, 2026

• CADE first blocked the terms in January. Meta got a court injunction to suspend that block, then CADE's tribunal reversed it again on March 4

• The ruling was unanimous; Meta's appeal was denied

• Italy's competition authority took a similar preventive measure, cited favorably by CADE’s rapporteur

• Meta has 5 days from official publication to restore mechanisms for excluded chatbots

• CADE's president called this the tribunal's first precedent on artificial intelligence

Why it Bites:

WhatsApp isn't just a messaging app in Brazil. It's infrastructure.

It's where millions of small businesses operate, where people pay bills, schedule appointments, and increasingly, where they interact with AI assistants.

Meta knows this.

The new terms weren't about security.

They were about owning the AI layer on top of the most dominant communication channel in the country.

What CADE did here is establish that controlling the pipe doesn't give you the right to control what runs through it.

At least not while the investigation is open.

That's a meaningful line.

The second-order effect is geographic:
Italy moved first, Brazil followed, and both cited each other.

This is how regulatory precedent spreads.

If the EU picks this up:

Meta's WhatsApp AI strategy gets a lot more complicated globally.

_{Everything Else}
🧠 You Need to Know

🤖 OpenAI Ships GPT-5.3 Instant, Calls Previous Version "Cringe"
→ Fewer refusals, no preachy preambles, 26.8% fewer hallucinations on web answers. GPT-5.4 already teased for this week.

🚫 QuitGPT Claims 1.5 Million Sign-Ups After Pentagon Deal
→ Online boycott surged after OpenAI signed with the DoD hours after Anthropic walked away, pushing users toward Claude, Gemini, and open-source alternatives.

🏛️ FTC Has Until March 11 to Say When Federal AI Law Beats State Rules
→ Trump's December executive order deadline lands next Tuesday. The answer will shape how 38 state AI laws survive or don't.

⚖️ Brazil's CADE Forces Meta to Keep Rival AIs on WhatsApp
→ Antitrust tribunal unanimously blocked Meta's new terms of use, keeping competing chatbots like Luzia and Zapia on WhatsApp Business while abuse-of-dominance investigations continue.

💡 NVIDIA Drops $4B on Photonics to Own AI's Next Wiring Layer
→ $2B each to Lumentum and Coherent, betting light-based interconnects replace copper as the backbone of next-gen AI data centers.