- The Shift
- Posts
- AI Just Breached a Crypto Vault
AI Just Breached a Crypto Vault
Plus, how to build a conversational analytics agent (No Code), Google has officially launched Gemini 3.1 Pro, and more!
In partnership with
Welcome back to The Shift. Let’s get straight to what matters in AI today…
Today we have:
💸 AI Can Now Drain a Crypto Wallet
📊 How to Build a Conversational Analytics Agent (No Code)
🚀 Gemini 3.1 Pro: Google’s Big Reasoning Leap
🔨Tools and Shifts you cannot miss
💸 AI Can Now Drain a Crypto Wallet
OpenAI and Paradigm just showed something that should make anyone in crypto pause. Its latest coding agent, GPT-5.3-Codex, can successfully exploit vulnerable smart contracts 72% of the time. And this wasn’t a stunt; it was measured in a controlled benchmark called EVMbench.
The Shift:
1. What EVMbench Actually Is - OpenAI and Paradigm built EVMbench to test whether AI can find (detect), fix (patch), and exploit (exploit) bugs in smart contracts, which are basically automated money vaults that hold over $100 billion in crypto assets. If the code has a flaw, the money can be drained instantly and permanently.
GPT-5.3-Codex scored 72.2% in exploit mode, up from GPT-5’s 31.9% just six months ago, and in one case, GPT-5.2 independently executed a complex flash loan attack that emptied a test vault in a single transaction without human guidance.
2. AI Is Still Worse at Defending - Detection and patching remain much harder, with top models catching under half of known vulnerabilities, though when given a small hint about where to look, patch success jumps from 39% to 94%.
OpenAI positions EVMbench as a cybersecurity tool and is backing that claim with $10 million in API credits for researchers, an expanded beta of its Aardvark security agent, and a Trusted Access program for vetted professionals.
TOGETHER WITH NEO
Trust-First AI, Built Into Your Browser
Agentic workflows are everywhere. Real trust is still rare.
Norton Neo is the world’s first AI-native browser designed from the ground up for safety, speed, and clarity. It brings AI directly into how you browse, search, and work without forcing you to prompt, manage, or babysit it.
Key Features:
Privacy and security are built into its DNA.
Tabs organize themselves intelligently.
A personal memory adapts to how you work over time.
This is zero-prompt productivity. AI that anticipates what you need next, so you can stay focused on doing real work instead of managing tools.
If agentic AI is the trend, Neo is the browser that makes it trustworthy.
Try Norton Neo and experience the future of browsing.
📊 How to Build a Conversational Analytics Agent (No Code)
Stop building dashboards. Start building agents.
Here’s how to use Thesys Agent Builder to create an AI analytics assistant that returns charts, tables, and reports instead of text:
1️⃣ Upload your data: Drop in a CSV file or connect your database. No pipelines. No setup headaches.
2️⃣ Ask questions in plain English
“Show revenue by customer segment.”
“Compare Q2 vs Q1 margins.”
“Create a slide summary for leadership.”
3️⃣ Get visual outputs instantly: Instead of paragraphs, your agent returns charts, tables, slide-ready summaries, and structured reports.
4️⃣ Share with your team: Send the link so anyone can explore data, ask follow-ups, and generate insights on demand.
No dashboards. No SQL. Just conversational analytics. You can claim your $10 free credits.
🚀 Gemini 3.1 Pro: Google’s Big Reasoning Leap
Google has officially launched Gemini 3.1 Pro, upgrading the core intelligence behind last week’s Deep Think update. It brings major reasoning gains while keeping API pricing unchanged. The rollout spans consumer apps, enterprise platforms, and developer tools.
The Shift:
1. Massive Reasoning Jump - Gemini 3.1 Pro scored 77.1% on ARC-AGI-2, more than doubling Gemini 3 Pro’s 31.1% and outperforming Opus 4.6 and GPT-5.2 on the same benchmark.
2. Built for Hard Problems - The model is designed for complex, multi-step reasoning across science, coding, agentic workflows, and structured logic tasks where shallow answers do not hold up.
3. Production-Ready Creativity - 3.1 Pro can generate website-ready animated SVGs directly from text prompts, producing clean, scalable code instead of pixel-based media files.
4. Broad Rollout, Same Pricing - Available across the Gemini app, NotebookLM, Vertex AI, and developer APIs, 3.1 Pro keeps identical pricing and a 1M token context window despite frontier-level performance.
Gemini 3.1 Pro signals Google’s push to compete at the very top of reasoning AI while holding the pricing line. For developers, researchers, and enterprises, it delivers stronger logic without forcing a cost tradeoff.
TOGETHER WITH PROTON
Privacy-first email. Built for real protection.
End-to-end encrypted, ad-free, and open-source. Proton Mail protects your inbox with zero data tracking.
🔨AI Tools for the Shift
🎛️ Moozix – Create music, analyze mixes, and generate promo videos in one AI-powered workflow that takes you from idea to release-ready output.
👗 LookCraft – Use AI to discover your best outfit combinations and refine your personal style with smart recommendations.
📞 Mio – Text your request, and Mio picks up the phone, handles the conversation, and reports back like your personal AI phone assistant.
🎼 Melodusk – Create any song you imagine in seconds with an AI music maker built for instant inspiration.
🎯 Braintrust AIR – An AI interviewer that reduces hiring costs, eliminates bias, and identifies top talent quickly, trusted by leading global companies.
🚀Quick Shifts
📱 Samsung’s new conversational Bixby has entered beta via One UI 8.5 in the US and select markets, following its TV rollout, with more details expected at next week’s Unpacked event.
💬 ChatGPT ads are now appearing in the wild, with brands like Expedia, Qualcomm, Best Buy, and Enterprise Mobility showing sponsored placements that can trigger as early as a user’s first prompt.
🛒 Reddit is testing an AI shopping search feature that turns community recommendations into interactive product carousels with pricing and buy links, blending organic discussion with ecommerce partnerships.
💰 OpenAI is reportedly finalizing a $100B funding deal at a valuation exceeding $850B, as major backers like Amazon, SoftBank, Nvidia, and Microsoft double down despite rising costs and new ChatGPT ads.
🔊 ElevenLabs secured the first insurance policy covering AI voice agents, earning certification that allows enterprises to insure AI-driven actions on its platform.
🧩Prompt of the Day
How to Plan Customer Education Content Using One Prompt
Turn scattered how-to content into a structured education plan that helps customers succeed and stay loyal.
Paste the prompt: Drop this into ChatGPT, then fill in your product and the main usage scenarios customers care about most.
Prompt to paste
Generate customer education content ideas for [Product]. Include key topics based on real usage scenarios, suggested formats (video, blog, infographic, email course), and distribution channels. Keep it concise, practical, and focused on helping customers get better results with the product.
That’s all for today’s edition see you tomorrow as we track down and get you all that matters in the daily AI Shift!
If you loved this edition let us know how much:
How good and useful was today's edition |
Forward it to your pal to give them a daily dose of the shift so they can 👇
Reply