LeCun, Silver, Goldie — Where The Best Minds In AI Are Going Next
Jason Dussault
Chief Executive Officer, Co-Founder
This Post is disseminated on behalf of Intellistake Technologies Corp.
I've been keeping a list. Names of researchers who've walked out of the biggest AI companies in the world over the last twelve months. This week the list jumped again, and it's worth talking about why.
Yann LeCun stepped down as Meta's chief AI scientist and closed a $1 billion round for AMI Labs. David Silver left Google DeepMind and just announced a $1.1 billion seed round for Ineffable Intelligence¹. Tim Rocktäschel walked out of DeepMind too, reportedly raising up to a billion of his own. Anna Goldie and Azalia Mirhoseini left Anthropic and Google to start Ricursive Intelligence, which has already pulled in $335 million. CNBC reports that venture capital has poured $18.8 billion into AI startups founded since the start of 2025… already on pace to outrun last year.
These aren't restless engineers chasing equity. These are some of the most decorated researchers in the history of AI, walking out of arguably the best jobs in technology. When the people closest to the center of a market start moving in the same direction at the same time, that's a signal. The interesting question is what the signal says.
The Race Got Smaller, Not Bigger
Elise Stern at Eurazeo, the venture firm that backed AMI Labs, gave CNBC the cleanest explanation I've read all year:
"When you're in a race, you narrow focus. That creates a vacuum. Entire areas of research, like new architectures, agents, interpretability and vertical models, are being deprioritised, not because they don't matter, but because they don't win the immediate race."
Read that twice, because it's the whole story.
The biggest AI companies in the world are now optimized for one contest. Bigger models. Faster releases. More impressive demos at the next keynote. That's the race they have to win, because the valuations they've raised demand it. Alexander Joël-Carbonell at HV Capital said the same thing in different words:
"Inside the large foundational labs, the pressure to deliver benchmark performance and maintain rapid release cycles leaves limited room for genuinely exploratory research, particularly outside the dominant LLM paradigm."
Look closely at what Stern listed in that vacuum. New architectures. Agents. Interpretability. Vertical models. Those aren't fringe topics. Those are entire categories of AI work that the biggest companies are choosing not to prioritize right now. Not because the work doesn't matter. Because it doesn't win the next keynote.That's the part most coverage is missing.
The Giants Narrowing Is Good News For Focused Companies
.jpg)
When the largest players in any market narrow their focus, they leave behind a much bigger map than they cover. And the companies best positioned to fill that map aren't the next set of giants. They're focused, applied companies built around solving real problems for real customers, not benchmarks.
Anna Goldie said something to CNBC that, for me, was the most revealing line of the whole article. Talking about why she had to leave Google to build a chip-design AI company, she said:
"For chipmakers to trust us with their most valuable IP, we have to be Switzerland, and that wouldn't be possible if we were at Google."
Think about what she's actually saying. Some of the most sophisticated technology buyers in the world won't hand their crown jewels to a vendor sitting inside Google. Not because the technology is bad. Because the relationship is conflicted. They want a focused, neutral, specialized partner. That isn't a niche. That's most of enterprise.
The Buyer Side Of The Same Gap
This week David Linthicum at InfoWorld² wrote a piece that was almost the mirror image of the CNBC story. Different angle, same gap.
His argument: enterprise AI adoption looks broad on the surface, but it's almost entirely His argument: enterprise AI adoption looks broad on the surface, but it's almost entirely happening at the edge of the business. Meeting summaries. Internal assistants. Document drafting. Useful, sure, but nowhere near the operational core of how a company actually runs. The systems that decide whether a business performs well or badly, like inventory, logistics, procurement, and financial transactions, are still mostly untouched by AI.
Solving the operational core of an enterprise is a different kind of problem than writing better email drafts. It needs vertical specialization. Real-world data. AI agents that can actually take action inside a workflow rather than just summarize one. And it needs partners who are willing to roll up their sleeves on the unglamorous parts of how a business actually works.
That kind of work doesn't come out of a frontier lab racing for the next benchmark. It comes from teams building close to the customer, in the categories the giants have decided not to prioritize.
Stern and Joël-Carbonell describe the supply side of the gap. Linthicum describes the demand side. They're staring at the same hole.
That hole is the entire opportunity for focused, applied AI companies right now.
The Bigger Shift Underneath All Of This
.jpg)
Step back from the talent moves for a second, and there's a larger pattern worth naming.
The same concentration that's pushing researchers out of the giants is pushing enterprise buyers in a new direction too. The biggest AI providers all run on a small handful of centralized cloud platforms. That worked fine when AI was an experiment. It works less well when AI is being asked to handle the operational core of a real business, with sensitive data that took years to build and competitive value that can't be casually parked inside a single vendor's environment.
That's why you're starting to see a real move toward decentralized and sovereign AI infrastructure. Not as a slogan. As a procurement decision. Enterprise buyers are asking harder questions about where their data physically sits, who has jurisdiction over it, whether the compute they're relying on is shared with everyone else or actually dedicated to them, and whether they're locked into one vendor's roadmap whether they like it or not.
You can see it happening in real procurement decisions. Intellistake just announced a strategic agreement with Singularity Compute³, the AI infrastructure arm of SingularityNET, for dedicated, sovereign AI compute capacity hosted in Sweden by a Swiss-incorporated provider. It's purpose-built for AI workloads, runs on 100% renewable energy, and sits outside the jurisdictional reach of the U.S. CLOUD Act.
I said something in that announcement I'll repeat here, because it's the same idea from a different angle:
"I believe there is a clear move towards decentralization, and this milestone sits within that broader shift in how AI infrastructure is being built and deployed. At its core, decentralization gives users more control over their data and assets, rather than relying on centralized systems where risk is concentrated and intermediaries sit between the user and the underlying infrastructure."
The talent moves and the infrastructure moves are the same story. The era when one or two giants could be the answer to everything in AI is ending. Sovereignty, specialization, and dedicated capacity are becoming features enterprise buyers actually pay for, not nice-to-haves.
One Simple Rule
Talent tells you a lot about where a field is heading. Researchers don't usually walk away from the best-resourced rooms in their industry unless they think the more interesting work is somewhere else. When several of them do it inside the same window, it's worth paying attention to what they're walking toward, not just what they're walking away from.
The giants will keep racing each other on benchmarks, and they'll keep doing that well. But the question that actually matters for the next decade of AI isn't who wins that race. It's whether AI ever makes it from impressive demo to operational reality inside the businesses that need it.
That answer isn't coming from inside a single keynote. It's being assembled, slowly, by a much wider set of builders working closer to the ground.
Disclaimer
There has been significant volatility in digital assets and their value can decline rapidly, which in turn would lead to a decline in the stock price of companies holding digital assets. Intellistake is a start-up that does not have the same access to capital as other larger more established companies.
Intellistake has just commenced operating its business and is at an early stage of development. Intellistake is entering this space by acquiring and operating blockchain validator hardware that supports AI networks and investing in AI-related digital tokens to primarily operate validator hardware.
Intellistake is presently evaluating the regulatory framework for tokenization. Any tokenization will be subject to it being completed in compliance with applicable law, regulatory requirements and terms of any underlying agreements associated with the underlying assets. The actual structure of such tokenization, the assets that would be subject to tokenization, and the associated timeline, have not yet been determined. Intellistake will provide further updates as material developments related to this tokenization strategy occur.
Intellistake is developing custom AI software systems called "AI Agents" for businesses. It recently announced the development of IntelliScope, a newly designed enterprise artificial-intelligence (AI) suite that applies decentralized AI technologies to deliver transparent and verifiable corporate intelligence. IntelliScope, which is in testing, is being publicly introduced as Intellistake's enterprise AI suite, reflecting the Company's focus on advancing practical applications of decentralized AI technologies.
The IntelliScope suite is being developed as a collection of modular AI agents, each intended to address specific enterprise challenges. Development has advanced through internal closed testing, where functionality is being refined and validated. Built to leverage decentralized AI technologies developed within the ASI Alliance FET token ecosystem, IntelliScope is now preparing to move into closed beta testing with an enterprise client, a phase focused on gathering feedback to shape premium features and expand real-world use cases.
The Company intends to deliver these solutions either as one-time projects or ongoing subscription services. Revenue is expected to come from implementation fees and monthly subscription payments. The Company does not presently have any customers. Intellistake is just commencing operations. It is targeting significant growth but its business is subject to several risks related to general business, economic and social uncertainties; the sufficiency of cash to meet liquidity needs; legislative, political and competitive developments; the inherent risks involved in the digital currency and general securities markets; the volatility of digital currency prices and the additional risks identified in the "Risk Factors" section of the Company’s filings with applicable securities regulators. Intellistake has not yet developed or commercialized its AI solutions.
Completion of the Singularity Venture Hub (“SVH”) acquisition remains subject to completion of satisfactory due diligence, the negotiation, and execution of a definitive agreement ("Definitive Agreement") that will include representations, warranties, covenants, indemnities, termination rights, and other provisions customary for a transaction of this nature, no objection from the Canadian Securities Exchange, and shareholder approval of SVH, if required.
This report contains "forward-looking information" concerning anticipated developments and events related to the Company that may occur in the future. Forward looking information contained in this report includes, but is not limited to, all statements in respect of the Company's growth and development, the operations and business segments of the Company, support for decentralized AI and blockchain networks, the details of the collaboration with Orbit AI and its expected benefits; the Company’s contributions towards the collaboration with Orbit AI; the timelines for Orbit AI’s operation; and Intellistake’s strategy to support tokenized, decentralized AI infrastructure.
In certain cases, forward-looking information can be identified by the use of words such as "expects", "intends", "anticipates" or variations of such words and phrases or state that certain actions, events or results "may", "would", or "might" suggesting future outcomes, or other expectations, assumptions, intentions or statements about future events or performance. Forward-looking information contained in this report is based on certain assumptions regarding, among other things, the Company and SVH satisfy all conditions necessary to close the proposed transaction; the Company will continue to have access to financing until it achieves profitability; obtaining the necessary regulatory approvals; the technology and blockchain industries in which the Company intends to focus its business in will grow at the rate and in the manner expected; the ability to attract qualified personnel; the success of market initiatives and the ability to grow brand awareness; the ability to distribute Company's services; the Company creates strategies to mitigate risks associated with cryptocurrency price fluctuations; the Company and SVH remain compliant with all applicable laws and securities regulations and applicable licensing requirements; the Company engages and collaborates with local experts, as necessary, to address jurisdiction-specific matters and ensures compliance with foreign regulations to avoid penalties; the Company addresses any potential cybersecurity threats promptly and effectively; the ability of the Company to develop its technology, acquire customers and have revenue; the ability to successfully deploy the new business strategy as a result of the change of business. While the Company considers these assumptions to be reasonable, they may be incorrect.
Forward looking information involves known and unknown risks, uncertainties and other factors which may cause the actual results to be materially different from any future results expressed by the forward-looking information. Such factors include risks related to general business, economic and social uncertainties; failure of the Company and SVH to satisfy all conditions necessary to close the proposed transaction; failure to raise the capital necessary to fund its operations; inability to create strategies to mitigate the risks associated with cryptocurrency price fluctuations; the costs of regulation in the digital asset industries increase to the extent that the Company is no longer generating sufficient returns for shareholders; failure to promptly and effectively address cybersecurity threats; insufficient resources to maintain its operations on a competitive basis; and the actual costs, timing and future plans differs expectations; legislative, environmental and other judicial, regulatory, political and competitive developments; the inherent risks involved in the cryptocurrency and general securities markets; the Company may not be able to profitably liquidate its current digital currency inventory, or at all; a decline in digital currency prices may have a significant negative impact on the Company's operations; the Company's success may depend on the continued involvement of key personnel, including advisors, whose involvement cannot be guaranteed; institutional adoption of decentralized AI infrastructure remains uncertain and may not occur at the pace or scale anticipated; evolving regulatory frameworks, including those related to AI (such as Canada's proposed Artificial Intelligence and Data Act), may impose additional compliance burdens or restrict certain business activities; valuation figures are based on publicly available market data and internal assessments at the time of the referenced transactions and may not reflect current or future valuations; the volatility of digital currency prices; the inherent uncertainty of cost estimates and the potential for unexpected costs and expenses, currency fluctuations; regulatory restrictions, liability, competition, loss of key employees and other related risks and uncertainties; delay or failure to receive regulatory approvals; failure to attract qualified personnel, labour disputes; and the additional risks identified in the "Risk Factors" section of the Company's filings with applicable Canadian securities regulators.
Although the Company has attempted to identify factors that could cause actual results to differ materially from those described in forward-looking information, there may be other factors that cause results not to be as anticipated. Readers should not place undue reliance on forward-looking information. The forward-looking information is made as of the date of this report. Except as required by applicable securities laws, the Company does not undertake any obligation to publicly update forward-looking information.
