PRAV

Investor and Operator | | ex @EIF_EU
@newnexhq Institutional VC Syndication Platform.
@seinethq EU-UK-India GTM and Partnership Platform

[PRAV @pjo1729]

The AI Object Framework describes a structural shift in which humans increasingly exist within AI mediated systems in two roles: as participants who interact with the system, and as entities that can be observed, represented, modelled and predicted by it.

The AI era is moving rapidly.

Today, we are debating open weight versus closed models, local versus cloud AI, centralised versus edge intelligence, and countless other developments unfolding in real time. Companies, investors and governments are making bets, and many of those bets are hotly contested.

But I think it is equally important to take a view on a different question: Where is this going?

Rather than focusing only on what AI is today, I have been thinking about where these developments eventually converge and settle over the next five, or perhaps ten, years.

What happens when we look beyond the next model release, the next funding round or the next technology cycle?

What are the structural developments most likely to shape the direction of the AI era?

I believe thinking clearly about the direction of travel gives us a better framework for understanding technology, evaluating investments and making personal and professional decisions.

Here are my 12 structural predictions for the AI era.

These are not predictions about which company will win or what the next AI model will look like. They are my view of the deeper shifts that could shape the technological, economic and social environment we are moving towards.

1. AI will move from the cloud to the edge.

Today, much of the most powerful AI runs in centralised data centres. Over time, intelligence will increasingly run directly on phones, PCs, vehicles, robots, industrial equipment and other devices.

This shift is driven by cost, latency, privacy, resilience and the simple fact that intelligence becomes far more useful when it can operate directly where data is generated and actions take place.

The cloud will remain important, particularly for training and large scale computation. But intelligence will increasingly move closer to the point of use.

2. Open weight AI will take the lead in deployment.

The frontier of AI may remain concentrated among a relatively fewer number of companies with access to enormous amounts of capital, compute and data. But deployment is a different question.

Open weight models will evolve to take the lead in deployment where cost, customisation, privacy, sovereignty and local control matter. Companies and governments will not always want their intelligence layer to be dependent on a single external provider. They will increasingly want the flexibility to run, adapt and control models within their own infrastructure, whether in the cloud, on private infrastructure or at the edge.

3. AI agents will replace software interfaces.

For decades, humans have learned how to use software. We open an application, navigate menus, enter information and manually complete a sequence of actions.

That model will increasingly change. People will simply tell machines what they want. The AI agent will determine which tools to use, what information is required and how to execute the task. The agent becomes the interface.

We will move from learning software to expressing intent.

4. Intelligence will become abundant and cheap.

Reasoning, research, coding, design, translation, analysis and other cognitive capabilities will become dramatically cheaper and more widely available. This does not mean expertise disappears. It means the cost of accessing many forms of intelligence declines. As intelligence becomes abundant, scarcity shifts elsewhere: compute, energy, data, infrastructure, trust, distribution and human attention.

The economic question will increasingly become not whether intelligence is available, but who can combine it with scarce resources and execute most effectively.

5. More of life will become gamified, competitive and viral.

Prediction markets are an early example of a broader shift. Information is no longer simply consumed. People can take positions on it, compete around it, build reputations through it and be rewarded for being right. The same dynamics can spread across news, finance, forecasting, education, entertainment and professional activity. Prediction, competition, rankings, rewards and social distribution will increasingly become part of how systems generate engagement and participation.

The line between information, entertainment, competition and economics will become increasingly blurred.

6. AI will move from the digital world into the physical world.

The first major wave of AI is transforming digital work. The next wave moves into the physical world.

AI combined with sensors, robotics and edge computing will transform manufacturing, logistics, agriculture, healthcare, defence, transport and eventually everyday life. The economic impact of AI may ultimately be far greater when intelligence can not only generate information but also perceive and interact with the physical environment.

Software intelligence becomes physical capability.

7. Compute and energy will become strategic resources.

AI does not exist independently of physical infrastructure. It requires chips, data centres, electricity, cooling, networking and increasingly sophisticated supply chains. As AI becomes a foundational layer of the economy, compute capacity and energy availability become strategic resources. The competition for AI leadership will therefore not only be about models and algorithms. It will also be about who controls the infrastructure required to run them.

Chips, compute and energy could become defining sources of economic and geopolitical power.

8. The internet will become predominantly agent to agent.

Today, the internet is largely designed for humans. We search websites, compare products, fill in forms, make purchases and communicate with other people. That architecture will increasingly change.

AI agents will search, negotiate, compare, purchase, book, transact and communicate with other agents. Humans will increasingly define objectives rather than manually execute every step. The human readable internet becomes a machine readable economy. This has potentially profound implications for search, advertising, e commerce, payments, marketplaces and digital identity.

Increasingly, the customer on the internet will be an AI agent.

9. AI plus quantum computing will force a fundamental rewrite of computational capability and digital security.

AI is already accelerating discovery, design, simulation, optimisation and automation. Quantum computing could eventually unlock new computational capabilities for problems that are difficult or impractical using classical computing alone.

The combination could expand what digital systems are capable of doing. At the same time, quantum computing threatens major parts of today's cryptographic foundations, while AI dramatically increases the scale and sophistication of cyberattacks and automated exploitation.

The result will be a fundamental rewrite of both what digital systems can do and how they must be secured.

10. Human plus AI will become the new economic unit.

The most important consequence of AI may not be that machines simply replace humans. Instead, individuals and small teams will increasingly work alongside fleets of AI agents. A single person could have AI agents supporting research, sales, coding, operations, finance, marketing and execution. This changes the economics of scale. The gap between a person using AI effectively and one who does not could become enormous.

The basic unit of economic productivity may increasingly become: Human plus AI.

11. Everything will be monitored.

As intelligence moves to the edge and becomes embedded in more devices, monitoring will become increasingly pervasive. Devices, cameras, vehicles, transactions, communications, workplaces, homes and online behaviour will generate continuous streams of data. AI will not simply record this information. It will analyse, correlate and interpret it. The system will increasingly identify patterns that individual humans cannot see.

As sensing becomes more widespread, the boundary between the physical and digital world will become increasingly blurred.

12. Privacy will become the exception, not the default.

The most important privacy challenge of the AI era may not simply be data collection. It may be inference. AI will increasingly infer what you believe, want, fear, trust and are likely to do, even when you never explicitly disclose that information. Your behaviour can reveal information that you never consciously chose to share.

This fundamentally changes the privacy debate. The privacy question changes from “Who has my data?” to Who is allowed to infer things about me?”

That may be one of the defining political, legal and social questions of the AI era.

The reinforcing chain. These predictions are not independent. They reinforce each other.

Edge AI -> Ubiquitous Intelligence -> Ubiquitous Sensing -> Continuous Monitoring -> Prediction -> Autonomous Agents -> Agent to Agent Transactions -> Human + AI Economic Units -> Humans as Participants and Observable Entities Within The System

This creates a much larger structural shift. AI moves from being something humans simply use to becoming an increasingly pervasive operating layer through which people, businesses, machines and institutions interact.

The shift From Humans use systems To Humans participate in systems that can also observe, represent, model and predict them.

This does not mean humans necessarily lose agency. The structural shift is that humans are no longer only outside the system as users. They increasingly exist within it as participants and as entities the system can observe, model and predict.

The core questions therefore may not simply be: Will AI replace humans? A more important question is: How will humans exist within systems that increasingly understand, predict and interact with them? AOF is an attempt to answer this.

These 12 predictions are an attempt to think beyond the current debates and towards the direction in which the underlying technologies may eventually converge. The exact path may change. The winners and losers will certainly change. Some of these predictions may take five years. Others may take ten or longer. But understanding the direction of travel matters.

Because the decisions we make about technology, investments, businesses, careers and our personal lives will increasingly be shaped by the systems that emerge from it.

If you are building something or investing in something, the important question is not only whether it works today, but where it fits within these structural shifts and whether the direction of travel is working in its favour.

https://t.co/o2BkzI4PqS

#ai #predictions #venturecapital @karpathy

[PRAV @pjo1729]

The second fund will continue to make investments in early-stage companies across the AI ecosystem, the spokesperson said. The fund will look to make concentrated bets, aiming to back between eight and 10 companies each year, according to the spokesperson, who said the goal is to lead funding rounds. Check sizes will range from the low millions to as large as $50 million or $100 million for the right opportunity.

https://t.co/Skc4atQ39o

[PRAV @pjo1729]

Salesforce and Anthropic announced Tuesday a sweeping expansion of their partnership, called Claudeforce, that pushes the world's largest customer relationship management platform directly inside Claude - a tacit acknowledgment that the future of enterprise software may not involve enterprise software's own screens at all.

Under the hood, the architecture is deliberately simple. When a seller asks Claude to update a deal, Claude first reasons over its available skills — "kind of human-like instructions," as Stokes described them — to determine whether specific guidance exists for the task. If it finds a match, it reads the instructions and executes against Salesforce's MCP server, which inherits the user's existing permissions. "If you don't own that record, if you don't have permission to see that record, the MCP server doesn't either, and so you won't be able to read or write it," he said. For enterprise buyers, that may be the announcement's most important technical claim: nothing new to stand up, nothing to re-audit, nothing to configure account by account.

The value of Salesforce is not in our UI itself. It's not the application. The value of Salesforce is in the data and the metadata, the years worth of kind of encoded workflows and business practices that have been built up inside of Salesforce. What we're doing is we're taking that and we're exposing it to a new UI.

https://t.co/TJ6jfI8scN

[PRAV @pjo1729]

I am always puzzled by the argument "something cannot do this now, we are cool".

The question one has to always ask is, what can this something do in the future and is that still cool. If not, how to make it cool and what can we do to prepare for it.

[PRAV @pjo1729]

Another reason people underestimate AI is that analogies to the effects of past innovations are misleading. We have no experience with a technology that can be adopted quickly or that can think and move like a human. When the PC came along, it took twenty years to significantly change how we worked because the software had to be developed, the price had to come down, and people had to learn how to use the tools and incorporate them into their business processes. AI, on the other hand, runs on the devices we already have, and it uses natural language. We don’t have to adapt to it because it can adapt to us. It can watch the same training video that is used to train human workers and learn from existing data. https://t.co/fqPm5NbP9s #ai

[PRAV @pjo1729]

The maker of Claude is likely to tell investors its potential revenue opportunities are above $30 trillion, topping SpaceX’s $28.5 trillion estimate, according to people familiar with the matter.

Anthropic more than doubled its revenue to $11.6 billion in the second quarter. To put its more than $30 trillion vision in context, the 191 technology companies in the S&P 1500 brought in $2.4 trillion in revenue last year, according to FactSet.
https://t.co/OnDWCyPjTc

[PRAV @pjo1729]

Download Joebox - your browser in a box https://t.co/stVXWiu09L
available for Mac, Windows and Linux

[PRAV @pjo1729]

a) have the propensity to combine geo-politics+mood-on-the-ground+economics+behaviorial sciences+ psychology+a bit of finance
b) discipline
c) opportunity to do so. u need the dough to show it https://t.co/TfzgT740Cr

[PRAV @pjo1729]

Germany makes super good sense given mostly well-established traffic rules - after London and Tokyo

Alphabet’s Waymo on Tuesday said it plans to roll out robotaxi rides in Germany in 2027, marking its third international market and first in the EU https://t.co/DbRAAFZlLt

[PRAV @pjo1729]

This matches the AI Object Framework predictions I posted some days ago "Open weight AI will take the lead in deployment."

"Thomson, the company's first proprietary large language model, developed in-house. Frontier labs have typically spent billions of dollars on compute and years of infrastructure investment to reach the frontier. Thomson Reuters took a different path: starting from a strong open-source foundation and investing $40 million to train Thomson into the right intelligence for the jobs that matter most, covering talent and compute. The result is a model Thomson Reuters fully controls, without the heavy inference costs of typical frontier models."

https://t.co/Zm8bWxTbrK

[PRAV @pjo1729]

https://t.co/ohcCwOe9KS

[PRAV @pjo1729]

Increasing costs for small businesses and entrepreneurs is a sure shot way to make big businesses become more powerful. Tons of never ending regulations to keep-up. Small entrepreneurs should become an integral part of drafting regulations.
https://t.co/jMiIcrgXgD

[Elso Labs - Tools that make complex things simple. @ElsoLabs]

UC Berkeley has open-sourced FreeToken, a new local inference engine purpose-built for Mixture-of-Experts (MoE) large language models. The project claims it delivers 2–4x faster inference speeds

https://t.co/SAmiJBlujt

[Elso Labs - Tools that make complex things simple. @ElsoLabs]

The main idea is that AI chips are no longer just about making the processor faster. The bigger challenge is moving data between memory, compute units, and other chips quickly and efficiently.

https://t.co/ngHPNOuTHm

#AI #architecture

[PRAV @pjo1729]

@shiftshapr @GavinSBaker yes important to focus on a) how humans can continue to have agency b) how to use natural resources efficiently and create prosperity - all the best on the project

[PRAV @pjo1729]

Thanks to @GavinSBaker for starting a real discussion about a real topic.

I cannot stop thinking that we may be debating the trees while forgetting the forest. Before we get into the what and how, perhaps we should first define the where:

What kind of world do we want to build with AI?

Discussions about whether AI should be concentrated or distributed, open or closed, regulated or unrestricted should ultimately be grounded in the kind of future we want AI to help create.

For me, there are five foundational principles:

1. People should retain agency over their own lives.
AI should expand human agency, not quietly erode it. People should be able to understand, choose, override and control systems acting on their behalf.

2. People should have meaningful access to AI - and meaningful control over how they use it.
If AI becomes a fundamental layer of economic and intellectual infrastructure, access to powerful AI should not be limited to a small number of firms or institutions. People should be able to choose, switch and shape the AI that works for them.

3. People should have economic opportunities - and the ability to scale them.
AI should not only make existing companies more productive. It should dramatically expand the number of people who can create, build businesses, own assets and participate in economic growth.

4. Human autonomy over our bodies and minds should remain fundamental.
AI will increasingly be able to predict, persuade, personalise and influence behaviour. We need to think carefully about where technology ends and individual autonomy begins, particularly around health, bodily decisions and cognitive freedom.

5. We should avoid unaccountable concentrations of power.
Whether corporate, governmental or technological, systems that give a small group disproportionate control over information, economic opportunity or individual choice create risks that deserve serious attention.

The foundational question therefore isn't “How do we make AI safe?” It is “How do we build an AI-enabled world in which people retain agency, have access to powerful AI, can participate in and benefit from economic growth, retain meaningful autonomy, and avoid becoming dependent on unaccountable concentrations of power?”

I actually think this is in everyone's interest - including those who already have significant wealth and power.

A prosperous society is more valuable when it is also stable, free, creative and broadly optimistic about the future.

Once we agree on the world we want, regulation, technology design and AI governance become questions of how to get there. Perhaps that is where the conversation should begin.

#AI

[Paul Graham @paulg]

Someone asked what I'd do if I were 17. I'd learn how to build LLMs from scratch, and then train ones as powerful as I could with whatever hardware I could get access to.

[PRAV @pjo1729]

The Golden State has attracted $366 billion, or around 90% of all the venture capital for the U.S. so far this year.

No other state came close, with second-ranked New York attracting one-thirteenth of the capital, with $27 billion in deals announced.

The state’s economy grew 5% last year to a record $4.25 trillion, making it larger than every country other than the U.S., China and Germany. It is home to nearly 400 billion-dollar startups - more than any other state.

https://t.co/Vv88LepmyP