Alan Kay

TCP/IP was needed to make consistent objects across the Internet. It has been done since, but not as a standard layer on tcp/ip.

Licklider funded the Internet and called it the intergalactic network to at least get a network that would reach every human being. It is a political system because there is no center. People will do things alien to us!

Sketchpad

Constraints and problem-solving functions are based on what the user tries to do. Masters and Instances. First system with objects and classes.

The system could be used to build bridges without knowing in advance about bridges.

McLuhan: most people experience the present in terms of the past

How to design smart ideas to

Compositing for User Interface

Pen Stroking

Texturing

Bitmap Filters

https://youtu.be/ubaX1Smg6pY?is=THe0FLdFrpM4D10A

Alan Kay on Schemas

Alan Kay—one of the pioneers of object-oriented computing and the creator of Smalltalk—generally had a skeptical view of rigid schemas in computing systems. His perspective comes from a broader philosophy about how software systems should evolve and communicate. Here are the key ideas behind his thinking.

From Julien Bek:

https://x.com/JulienBek/status/2029680516568600933

The next $1T company will be a software company masquerading as a services firm. Every founder building an AI tool is asking the same question: what happens when the next version of Claude makes my product a feature? They’re right to worry. If you sell the tool, you’re in a race against the model. But if you sell the work, every improvement in the model makes your service faster, cheaper, and harder to compete with. A company might spend $10K a year for QuickBooks and $120K on an accountant to close the books. The next legendary company will just close the books.

Intelligence vs Judgement

Writing code is mostly intelligence. Knowing what to build next is judgement.

Translating a spec into code, testing, debugging: the rules are complex but they are rules. Judgement is different. It requires experience and taste, instinct built on years of practice. Deciding which feature to build next, whether to take on tech debt, when to ship before it’s ready.

A year ago, most Cursor users treated AI as autocomplete. Today, more tasks are started by agents than by humans. Software engineering accounts for over half of all AI tool usage across professions. Every other category is still in single digits. The reason is that software engineering is primarily intelligence work. AI has crossed the threshold where it can do most of the intelligence work autonomously and leave the judgement to humans. Software engineering got there first. It is coming to every single profession.

Source: Anthropic

Copilots and Autopilots

A copilot sells the tool. An autopilot sells the work.

Until recently, AI models were still developing intelligence and judgement, so the right approach was to build a copilot first: put AI in the hands of a professional and let them decide what to do with it. Harvey sells to law firms. Rogo sells to investment banks. The professional is the customer, the tool makes them more productive, and they take responsibility for the output.

Today, the models are intelligent enough that in some categories the best place to start is as an autopilot. Crosby sells to the company that needs an NDA drafted, not to outside counsel. WithCoverage sells to the CFO who needs insurance, not to the broker. The customer is buying the outcome directly. The work budget in any profession dwarfs the tool budget, and autopilots capture the work budget from day one.

The higher the intelligence ratio in any field, the sooner autopilots will win.

The Convergence

Today’s judgement will become tomorrow’s intelligence. As AI systems accumulate proprietary data about what good judgement looks like in their domain, the frontier will shift. Copilots and autopilots will converge. The copilot-to-autopilot transition has already begun in several categories. But the starting position matters because it determines where autopilots can win customers now and begin compounding the data that will eventually let them handle judgement too.

The Autopilot Playbook: Outsourcing as the Wedge

For every dollar spent on software, six are spent on services.

The total addressable market for autopilots is all labour spend in a category, insourced and outsourced combined. But the right place to start is where outsourcing already exists.

If a task is already outsourced, it tells you three things. One, the company has accepted that this work can be done externally. Two, there’s an existing budget line that can be substituted cleanly. Three, the buyer is already purchasing an outcome. Replacing an outsourcing contract with an AI-native services provider is a vendor swap. Replacing headcount is a reorg.

The playbook: companies should start with the outsourced, intelligence-heavy task. Nail distribution. Expand toward the insourced, judgement-heavy work as the AI compounds. The outsourced task is the wedge. The insourced work is the long-term TAM.

Crosby started with NDAs: a well-defined task, primarily intelligence, that most companies already outsource to external counsel. The budget exists, the scope is clear, the ROI is immediate, and the substitution is frictionless.

Opportunity Map

Plotting every service's vertical on an intelligence-to-judgement spectrum and outsourced-to-insourced ratio produces a priority map with labour TAM in brackets. The list is illustrative.

Source: Sequoia Research

Insurance brokerage ($140-200B). The largest dollar market on this list. Standard commercial lines are highly standardised: the broker’s value-add is essentially shopping across carriers and filling forms, pure intelligence work. The distribution layer is incredibly fragmented, tens of thousands of small brokers each running the same process, so no single incumbent controls the customer relationship. WithCoverage and Harper are interesting newcomers.

Accounting and audit ($50-80B outsourced in the US alone). The US has lost roughly 340,000 accountants over five years while demand has grown. 75% of CPAs are nearing retirement, the licensing path is long, and starting salaries lag tech and finance. That structural shortage is pushing firms to accept AI faster than almost any other profession. Rillet is building the AI-native ERP that will close the books. Basis started as a copilot for accountants.

Healthcare revenue cycle ($50-80B outsourced in US). People hear “healthcare” and assume it’s judgement-heavy, but the billing layer is almost pure intelligence. Medical coding is translating clinical notes into ~70,000 standardised ICD-10 codes. The rules are complex but they are rules. The outsourcing is already mature and outcome-based. An autopilot just has to do the same thing at lower cost. Anterior is the furthest along.

Claims adjusting ($50-80B including TPAs). On the other side of the insurance policy, claims adjusting is a separate autopilot surface. Standard-line claims are settled by interpreting policy language against damage schedules and setting reserves using actuarial tables. The adjuster workforce is aging out and nobody’s replacing them. The market is massively outsourced to independents and TPAs like Crawford and Sedgwick. One industry, at least two distinct autopilot opportunities. Pace is building the autopilot for claims handling. Strala is building an AI-native TPA.

Tax advisory ($30-35B). CPA licensing creates a regulatory moat, but 80-90% of the underlying work is intelligence. Every additional jurisdiction a tax autopilot handles deepens its data moat. Multi-jurisdiction complexity is exactly what SMBs outsource because no single in-house accountant can cover it. TaxGPT is an early mover alongside Skalar and Ravical in Europe.

Legal, transactional work ($20-25B). Contract drafting, NDAs, regulatory filings: high intelligence, routinely outsourced. The work product is standardised enough that quality is verifiable, so the buyer can trust AI output without deep legal expertise. Harvey is the emerging leader and is moving quickly to autopilot; Crosby and Lawhive are the autopilot-native newcomers.

IT managed services ($100B+). Every SMB outsources its IT. Patching, monitoring, user provisioning, alert triage: intelligence work running on repeat across thousands of identical environments. The existing software layer (ConnectWise, Datto) sells tools to the MSP. Nobody has yet sold “your IT runs” directly to the company as an outcome. Edra is automating IT processes. Serval is automating IT support.

Supply chain and procurement ($200B+). Most enterprises negotiate seriously with only their top 20% of suppliers. The long tail gets zero attention because it’s not economical to have humans do the work. Contract leakage runs 2-5% of total procurement spend. The wedge is abandoned work: no budget line to justify, no incumbent to displace, just found money. Magentic is building the AI for direct procurement, AskLio for indirect procurement. Tacto is building both the system of record and copilot for the midmarket.

Recruitment and staffing ($200B+). The largest services market on this list. The top of the hiring funnel (screening, matching, outreach) is pure intelligence, but closing a candidate and assessing culture fit is judgement built on years of pattern recognition. The autopilot wedge exists in high-volume, low-judgement roles where matching is standardised. Juicebox, Mercor, Jack & Jill are emerging winners building across the spectrum.

Management consulting ($300-400B). Huge market but the work is mostly judgement. The interesting question is whether AI can disaggregate consulting into intelligence components (data gathering, benchmarking) and judgement components (strategic recommendations), with the intelligence layer getting automated and the judgement layer staying human. Best candidates TBD.

In 2025, the fastest-growing AI companies were copilots. In 2026, many will try to become autopilots. They have the product and the customer knowledge. But they also face the innovator’s dilemma: selling the work means cutting their own customers out of doing it. That’s the opening for pure-play autopilots.

If you’re building one, reach out.

@julienbek

julien@sequoiacap.com

Real Objects

Real Objects, as in Alan Kay, not as encapsulation, abstraction, polymorphism, and inheritance. Most modern “OOP” (Java, C++, etc.) Focus on Class Types Inheritance But Kay’s original idea was to focus on Messaging, Independent objects, Late binding, and Extensible systems. Think about objects and messages, and not in terms of Remote Procedure Call. We are missing key properties in our current Objects (Notifications, Vault, Contacts, Linkbase). Because of this, our systems tend to become brittle, centralized, and difficult to evolve. Objects are behaviours and states. Objects can be regular objects or Agents. An object is like a small computer. Inspired by Alan Kay, we should aim for objects that can operate at internet scale—without central coordination, and without embedding infrastructure concerns into their design. Real Object Pillars We are still missing foundational building blocks. to make the system work But what matters now is having a clear direction—so each step moves us closer to a system that is composable, decentralized, and alive. Real Objects are the structure that LLMs and AI want.