Ilya Maps The Path Beyond AGI

Hello there! Ready to dive into another upgrading, Mind-boggling, and Value-filled Shift?

Today we have:

🤖 Ilya Sutskever’s Big Ideas: The Future of AI, Learning, and Research

🎬 How to Create Promo Videos in Minutes with Veed AI

🚀 AI Assistants With Memory Arrive in Perplexity

🔨Tools and Shifts you cannot miss

🤖 Ilya Sutskever’s Big Ideas: The Future of AI, Learning, and Research

Ilya Sutskever’s latest interview with Dwarkesh Patel lays out why today’s AI feels powerful yet brittle, why scaling is nearing its limits, and why a new research-driven era is emerging. He explains human learning advantages, emotional decision-making, and why the real goal is superintelligent learners, not finished AGI. 

The Shift:

1. Jagged Models - Ilya says current models are “jagged,” exceptional on PhD-level evals yet unreliable on simple tasks. Like overtrained test-takers, they lack true learning depth, revealing brittle generalization that scaling alone cannot fix.

2. End of Scaling - He argues pre-training is hitting data limits, and AI is returning to an “age of research.” With more companies than ideas, compute is no longer the differentiator; new learning principles matter more than bigger models.

3. Missing Principle  - Humans can master complex skills with tiny amounts of experience, like learning to drive in 10 hours, while AI needs millions of simulated attempts. Ilya says this gap comes from a “missing ML principle” that current AI doesn’t use, and he strongly hints SSI is trying to discover and implement it, though he refuses to reveal what it is.

4. Super Learners - Ilya says AGI won’t be a finished oracle but a superintelligent learner, like a brilliant 15-year-old scaling through deployment and shared experience. He stresses a gradual rollout so society feels AI capability, projecting 5–20 years.

Ilya’s vision centers on deeper learning principles, human-like adaptability, and elegant research over brute force. Progress requires superintelligent learners, not maximal scaling, with deployment guiding alignment, behavior, and society’s understanding of AI power.

^{TOGETHER WITH SYNTHFLOW}

A Better Way to Deploy Voice AI at Scale

Most Voice AI deployments fail for the same reasons: unclear logic, limited testing tools, unpredictable latency, and no systematic way to improve after launch.

The BELL Framework solves this with a repeatable lifecycle — Build, Evaluate, Launch, Learn — built for enterprise-grade call environments.

See how leading teams are using BELL to deploy faster and operate with confidence.

Get the Guide

🎬 How to Create Promo Videos in Minutes with Veed AI

Veed AI makes video creation simple, fast, and scalable, perfect for brand marketers who want to churn out reels, ads, or social content without messy editing.

1. Upload or record your asset - Drop in raw footage, product clips, or screen recordings; even a simple smartphone shot works.

2. Use Veed’s AI tools - Let Veed auto-edit: trim, subtitle, add captions, music, and transitions with just a few clicks. No editing skills required.

3. Brand & style it fast - Apply consistent brand colors, typography, and layout across videos, ideal for fast ad creatives or recurring content.

4. Export for any platform - Get final edits in formats ready for Meta, TikTok, Reels, or email, perfect for cross-channel campaigns.

✨ Great for teams and brands who need to produce lots of video content quickly and cleanly.

Try Veed → https://www.veed.io/ 

🚀 AI Assistants With Memory Arrive in Perplexity

Perplexity has introduced persistent memory, giving Comet Assistant the ability to remember preferences, past conversations, and user context automatically. This eliminates repetitive prompting and makes interactions faster and more accurate. The result is a more powerful, personalized assistant that adapts continually to how you think and work.

The Shift:

1. Smarter Long-Term Personalization - Perplexity now stores structured preferences like favorite brands, dietary rules, recurring topics, and work habits, allowing the assistant to maintain continuity across sessions without re-explaining details. 

2. Retrieval Instead of Guessing - Instead of treating user history as generic training data, Perplexity retrieves stored memory directly during generation, producing more precise recommendations, follow-ups, and reasoning grounded in your actual context. 

3. Higher Precision for Complex Tasks - Memory strengthens the assistant’s ability to handle layered queries: gift ideas, trip planning, book suggestions, or recalling prior advice. It uses past interactions to enhance reasoning for multi-step tasks. 

4. Full Control and Cross-Model Portability - Users can toggle memory on or off, delete items, and use incognito mode where nothing is saved. All data is encrypted and fully user-controlled. Memory also works across every Perplexity model, preserving your context no matter which assistant you switch to.

Memory transforms Perplexity from a one-off query tool to a persistent second brain that carries context across tasks. It removes friction, boosts productivity, and protects your accumulated knowledge as new models appear. This is a major step toward truly personalized AI assistance.

^{TOGETHER WITH GALACTIC FED}

This 20-minute growth session could save you millions.

Looking for growth answers in all the wrong places?

Now's your chance to meet 1-1 with Galactic Fed's advisors. They will understand your business, audit your channels, and show you where the fastest growth opportunities are hiding.

Their team’s data comes from managing millions in ad spend across 600+ brands like Quizno's, Varo, Edible and more. And they'll apply all that insight to your marketing.

It’s free, it’s personalized, and it’s the same process that’s helped clients get funded, acquired, and go from startup to standout.

Book your free expert session today →

🔨AI Tools for the Shift

🎨 PoYo.ai - Access premium AI models for image, video, music, and chat generation built for developers who need speed, quality, and affordability.

🗺️ GMapsScraper AI - Extract thousands of verified business leads from Google Maps in minutes with phone numbers, emails, addresses, and instant Excel export.

🎬 LetsMkVideo - Turn text and photos into professional AI videos with captivating scenes, lifelike motion, and stunning effects in seconds.

✉️ Touchpoint - Create beautiful emails and campaigns without coding or design skills using a simple AI-powered editor.

📖 Talk To Myra - Record short voice chats and instantly receive polished written audio and visual keepsakes that preserve your stories.

🚀Quick Shifts

🤖 China warns a humanoid-robot bubble may be forming, noting 150+ companies rushing in and many producing “highly similar” models. Officials say investment is outpacing real use cases, raising risks as R&D funding tightens.

🤓 Alibaba enters the AI smart-glasses race with Quark, featuring swappable batteries. The new S1 and G1 models offer cameras, bone-conduction mics, micro-OLED displays, and a dual-battery system delivering up to 24 hours of use, starting at 1,899–3,799 yuan.

🤖 Bridgewater’s Greg Jensen says the real AI bubble hasn’t even started. He warns investors “have no idea what’s hitting them,” noting AI already drives 1% of US GDP growth, while a global “resource grab” for chips, land, power, and scarce talent is accelerating.

🤖 Jeff Bezos’ stealth AI effort “Project Prometheus” has quietly acquired General Agents, a fast-growing agentic computing startup, according to Wired. 

📚 OpenAI just lost a major discovery ruling, the company must now turn over internal communications explaining why it deleted two datasets containing allegedly pirated books.

🤳AI Nugget of the Day

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:

Forward it to your pal to give them a daily dose of the shift so they can 👇