In partnership with
OpenAI just dropped a cybersecurity model that can talk about exploits and malware without flinching. GPT-5.4-Cyber breaks the usual safety rules on purpose so defenders can finally use AI to fight back. Basically to counter Claude Mythos
But wait, there's more:
Snap just fired 1,000 people and blamed AI for making them "too efficient"
Google's new voice model speaks 70+ languages and lets you control exactly how it sounds
I'm Alex. Welcome to L8R by Innov8.
Let's dive deep 🐰
In today's post:
OpenAI ships GPT-5.4-Cyber, a model built to hack back
Snap cuts 1,000 jobs as AI does their work
Google's Gemini 3.1 Flash TTS speaks 70+ languages with emotion control
OpenAI drops a hacker-friendly AI model
OpenAI just built an AI model that breaks its own safety rules on purpose.
GPT-5.4-Cyber is a massive new tool designed specifically to help cybersecurity defenders.
It lets security teams analyze malware and find software bugs without the AI refusing to help.
The details:
This special version of GPT-5.4 skips normal safety filters so it can read and reverse-engineer malicious computer code.
You cannot just log in and use it. Access is locked behind OpenAI's Trusted Access for Cyber program, which requires strict identity checks.
Top-tier users must let OpenAI monitor their chats to ensure they are not using the tool to launch real cyberattacks.
The rollout started on April 14, 2026, and aims to reach thousands of verified security experts and hundreds of teams.
Why it matters:
AI companies are fighting over how to release models that can be used as weapons. Rival company Anthropic recently gave its powerful Claude Mythos model to only eleven organizations because of extreme risk concerns. OpenAI thinks locking up AI is a mistake. They want to give GPT-5.4-Cyber to thousands of defenders so smaller companies are not left helpless against hackers.
💡 L8R's Take:
OpenAI is making the right call here. If you only give god-tier security AI to massive tech giants, the rest of the internet burns. We have to arm the good guys, even if the tools are dangerous.
From our Partner
Smart starts here.
You don't have to read everything — just the right thing. 1440's daily newsletter distills the day's biggest stories from 100+ sources into one quick, 5-minute read. It's the fastest way to stay sharp, sound informed, and actually understand what's happening in the world. Join 4.5 million readers who start their day the smart way.
Snap Fires 1,000 Workers Because of AI
Snap is dropping 16% of its global team to let AI do the heavy lifting.
CEO Evan Spiegel says new AI tools can crush repetitive work and speed things up.
This massive cut shows how fast AI is eating regular tech jobs.
The details:
Snap is laying off about 1,000 full-time workers and closing 300 open roles.
The company says AI agents now write at least 65% of their new code.
This move will save Snap over $500 million a year by late 2026.
Fired US workers get four months of severance pay and healthcare coverage.
Wall Street loved the news, pushing Snap's stock price up by over 6% right after the announcement.
Why it matters:
Tech companies are not just cutting costs anymore. They are completely changing how they work. Big names like Meta, Amazon, and now Snap are replacing human teams with AI agents. The era of hiring thousands of coders just to maintain apps is officially over.
💡 L8R's Take:
AI is no longer a fun toy for tech giants. It is a ruthless tool for boosting profits. If you write basic code or do repetitive tasks, you need to level up right now.
Google's New Gemini Voice Model Gives You Total Control
You can now direct AI voices exactly like human actors.
Google just dropped Gemini 3.1 Flash TTS, a massive audio model that speaks over 70 languages.
It completely changes the game by giving you precise control over how the AI speaks, pauses, and expresses emotion.
The details:
The model uses natural-language 'audio tags' so you can type commands to change the tone, pace, and delivery.
It supports native multi-speaker dialogue, letting you generate a full podcast with different voices in just one prompt.
The model scored a massive 1,211 Elo rating on the Artificial Analysis TTS leaderboard.
Google embeds SynthID watermarking into every audio clip to clearly identify it as AI-generated.
Developers can test the model right now in public preview through Google AI Studio and the Gemini API.
Why it matters:
AI voices used to be a total black box. You typed text and just hoped it sounded good. Now, you get a director's chair. This makes creating complex video voiceovers, localized content, and interactive media much cheaper. You do not need to stitch together dozens of audio files anymore.
🚀 Quick L8R Summary
GPT-5.4-Cyber: OpenAI just dropped a security-focused model that can talk about exploits and vulnerabilities without freaking out.
Snap Layoffs: Snap cut 1,000 jobs because AI made their teams faster efficiency gains hit different when you're on the wrong side.
Gemini Flash TTS: Google's new text-to-speech model speaks 70+ languages and lets you control tone with simple text tags.
📩 Innathe L8R engane undarunnu 👇?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha L8R il kanaam bie…👋