YC Partner Reveals: Building an AI-Native Company from Scratch

marsbitPublished on 2026-05-15Last updated on 2026-05-15

Abstract

"YC Partner Reveals: Building an AI-Native Company from Scratch" YC partner Diana Hu argues that true AI-native companies operate 1000x faster than incumbents, not by using AI for mere efficiency, but by making it the company's core operating system. This requires a fundamental shift: companies must become "queryable" to AI, with all workflows and communications generating data for AI to learn from, creating a "closed-loop" system for continuous optimization. For example, an AI agent with access to tickets, code, meetings, and customer feedback can analyze past performance and autonomously plan future engineering cycles, dramatically increasing output. In product development, the new paradigm is the "AI software factory": humans write specifications and tests, while AI agents generate the code. This transparent, data-driven model renders traditional middle management obsolete. Future AI-native companies will consist of three roles: Independent Contributors (who build/operate with AI), Directly Responsible Individuals (who own outcomes), and the AI Founder who leads by example. The critical shift is maximizing token usage over headcount. A small, AI-augmented team can outperform large traditional teams. Startups have a key advantage: they can design their entire culture and systems around AI from day one, unburdened by legacy processes. The core takeaway: Founders must personally experience AI's transformative power. The future belongs to those who embed AI into their com...

Source: Y Combinator

In Silicon Valley, Y Combinator (YC) is widely recognized as the "philosopher's stone" for global startups.

As the world's premier startup incubator, since its founding in 2005, YC has incubated over 5,600 companies, giving birth to tech giants such as Airbnb, Stripe, Dropbox, Reddit, and Coinbase. Even OpenAI's CEO Sam Altman once served as YC's President.

It can be said that YC's perspective represents the cutting-edge trends in tech entrepreneurship. Recently, YC partner Diana Hu proposed a striking assertion in the podcast "How To Build A Company With AI From The Ground Up": The operating speed of an AI-native startup could be up to 1000 times faster than existing industry giants.

TinTinLand has compiled the key insights from the original video. Let's explore, from YC's perspective, how a truly AI-native company should operate.

Not "Using AI," but "Running on AI"

Currently, most discussions about AI still revolve around "increasing efficiency," such as "AI can make engineers more efficient" or "we need to add a Copilot to our existing workflow." This mindset is fundamentally misguided.

The true transformation is not about productivity gains, but the emergence of entirely new capabilities.

A true AI-native company should not merely treat AI as a tool, but rather see it as the company's operating system (OS). In this model, every workflow, every decision, and every process should be handled through an intelligent layer that continuously learns and improves.

With the support of AI tools, a single capable individual can now build features that previously required an entire team, or even things that were previously impossible.

Making the Entire Company AI-Queryable

Building a Closed-Loop System

Diana introduced the "Closed Loop" concept from control systems to describe the ideal AI company.

  • Open Loop System: This is how traditional companies operate. Management makes decisions, employees execute, but results often cannot be measured and fed back systematically, with significant information loss throughout the process.

  • Closed Loop System: The system continuously monitors outputs, captures information, and feeds it back to the AI, thereby constantly optimizing processes over time.

The Prerequisite for a Closed Loop: Queryability

To achieve such a closed loop, it is necessary to make the company completely transparent and queryable to AI.

This means that all actions within the company must produce "digital artifacts" that AI can learn from:

👉 Use AI assistants to record meetings throughout, reduce the use of private messages and emails, embed AI agents in all communication channels, and build a company-wide real-time dashboard covering revenue, sales, engineering, hiring, and operations.

Specific Case: Revolutionizing Engineering Management

Diana gave a concrete example from engineering management: Suppose you have an AI Agent with access to Linear tickets, Slack channels, GitHub repositories, Notion docs, customer feedback emails, and daily standup meeting recordings.

Then, this Agent can genuinely analyze what was actually delivered in the last sprint and how well it matched customer needs—rather than relying on the distorted information from multiple layers of reporting.

Building on this, the Agent can go a step further: automatically propose engineering plans for the next sprint, making them more predictable and accurate. Diana mentioned that she has seen teams adopting this approach halve their engineering timelines while accomplishing nearly ten times more work.

The core principle behind this is: To gain the full capabilities of AI, you need to provide the model with the same level of context as your employees.

Software Factory: Humans Define Specs, AI Writes Code

At the product development level, a new paradigm is emerging—the AI Software Factory. This is an evolution of Test-Driven Development (TDD):

  • Humans Define Success: Humans write requirement specifications (Specs) and test cases defining success criteria.

  • AI Handles Implementation: AI Agents generate the code implementation and iterate until all tests pass.

  • Shift in Human Role: Humans define what to build and judge the output; writing the code itself is the Agent's job.

Diana noted that some leading companies have already achieved codebases with no handwritten code at all, only Specs and test suites.

This is also how the "10x engineer" envisioned by software engineer Steve Yegge can be realized: surround a single engineer with a systematic cluster of Agents, enabling them to build things that would have been impossible to achieve alone before.

Flat 2.0: A New Organizational Structure

When a company becomes queryable and information flow becomes transparent and driven by an AI layer, the traditional pyramid management structure becomes obsolete.

Traditionally, we needed middle managers to convey information up and down the organization. But in the new world, the AI intelligent layer takes on this role. If your company is queryable and highly digitized, you should need almost no "human middleware."

Every layer of human routing you eliminate is a direct speed gain.

The Three Types of Employees in Future Companies

Diana referenced a point by Block (formerly Square) founder Jack Dorsey: If you keep your old organizational structure and management practices, you're completely missing out on this wave.

Future AI-native companies will consist of the following three types of employees:

  • Type 1: Independent Contributors (ICs). These are the people directly creating and operating things. In an AI-native company, this isn't limited to engineers—operations, support, sales—everyone brings a working prototype to meetings, not just a PowerPoint.

  • Type 2: Directly Responsible Individuals (DRIs), focusing on strategy and customer outcomes. This isn't a manager in the traditional sense, but someone with clear accountability for a specific result.

  • Type 3: AI Founders, at the forefront, leading by example to show the team the capability gains brought by AI, not delegating AI strategy to others.

Key Shift: Maximizing Token Usage

👉 The most critical shift for an AI-native company is not maximizing headcount, but maximizing Token usage.

  • Leaner Teams: A single employee working with AI tools can produce output equivalent to what previously required a large engineering team.

  • Restructured Budget: Founders should be willing to pay very high API bills. Because these bills replace extremely expensive and bloated personnel costs.

In this model, startups can generate enormous impact with a very small scale.

The "Dimensionality Reduction" Advantage for Startups

Why is now the best time for startups to surpass giants?

Diana pointed out that large incumbent companies face serious "path dependency." They must maintain existing businesses while undoing years of accumulated standard operating procedures (SOPs) and core assumptions. For them, changing core processes is extremely risky.

In contrast, AI-native startups hold a massive advantage:

You can design the entire system, workflow, and company culture around AI from day one. The result is that AI-native startups can operate potentially 1000 times faster than existing industry giants.

Conclusion: The Non-Outsourcable Belief

Finally, Diana offered a crucial admonition: Do not outsource your belief in the power of AI tools; you must experience it firsthand yourself.

You must personally sit at the computer and work with programming Agents until you witness with your own eyes how they shatter your perception of "what is possible."

For early-stage founders, this is the best of times: no legacy systems to bind you, no thousand-person team to retrain, no entrenched organizational structure. You have the freedom to build the company right from the start.

The future winners will be those who dare to embed AI into the soul of their company from day one.

Related Questions

QWhat is the core difference between an AI-native company and a traditional company using AI tools, according to the article?

AThe core difference is that a true AI-native company treats AI as the operating system (OS) of the entire company, integrating it into every workflow, decision, and process as a continuous learning intelligent layer. In contrast, traditional companies often treat AI merely as a tool for improving efficiency ('giving our process a Copilot').

QWhat does Diana Hu mean by a 'Closed Loop' system in an AI-native company?

AA 'Closed Loop' system is one that continuously monitors outputs, captures information, and feeds it back to the AI, allowing the processes to be optimized over time. This contrasts with an 'Open Loop' system typical of traditional companies, where decisions are made and executed but results are not systematically measured and fed back, leading to significant information loss.

QWhat key principle must be followed to achieve a full AI-powered closed-loop system within a company?

ATo achieve a full AI-powered closed-loop system, the company must be made completely transparent and queryable to the AI. This means all internal actions must produce 'digital artifacts' for the AI to learn from, and the AI must be provided with the same level of contextual information as human employees.

QHow does the article describe the concept of an 'AI Software Factory' and the changing role of engineers?

AThe 'AI Software Factory' is an evolution of Test-Driven Development (TDD). In this paradigm, humans define the specifications (Specs) and write test cases to define success criteria. AI Agents then generate the code to meet these specs and iterate until all tests pass. The engineer's role shifts from writing code to defining what to build and judging the output; the act of coding itself becomes the work of the AI Agent.

QAccording to the article, what is the fundamental budget and resource shift for an AI-native startup compared to a traditional one?

AThe fundamental shift is from maximizing headcount to maximizing token usage. Founders should be willing to pay very high API bills, as these costs replace the far more expensive and cumbersome costs of large human teams. A lean team empowered by AI tools can achieve output equivalent to a much larger traditional engineering team.

Related Reads

Can DeepSeek Save China One Trillion Dollars?

"DeepSeek and the $1 Trillion Infrastructure Question" The article examines whether DeepSeek's AI optimization breakthroughs could potentially save China $1 trillion in future AI infrastructure costs. The analysis begins with Nvidia's upcoming Vera Rubin AI platform, costing ~$7.8 million, where memory (HBM4/LPDDR5X) constitutes $2 million—a 435% cost increase in one year, highlighting how AI hardware spending is shifting toward expensive memory components. DeepSeek's approach works in the opposite direction. Through three key technical innovations showcased in DeepSeek V4, the company dramatically improves hardware efficiency: 1. **Memory Compression (MLA)**: Re-engineers the attention mechanism to compress long-context memory (KV Cache) by over 90%, drastically reducing expensive HBM usage. 2. **Selective Activation (MoE)**: Employs Mixture-of-Experts architecture where only a small fraction of parameters (e.g., 49B out of 1.6T in V4-Pro) are activated per token, allowing most parameters to reside in cheaper memory/SSD. 3. **Computation Caching**: Reuses previously computed results via cache hits, replacing expensive GPU computations with cheap memory reads. Combined, these optimizations allow the same hardware to produce approximately 4x more tokens, effectively reducing required hardware investment by 75%. DeepSeek's pricing reflects this: a 10-billion token workload costs ~$522 monthly versus ~$9,000-$10,000 for competitors. The $1 trillion savings projection stems from McKinsey's estimate that global AI infrastructure will require ~$5.2 trillion investment by 2030. As China's daily token consumption grows toward quadrillions, even marginal efficiency gains scale massively. With a conservative 4x throughput improvement, China could avoid building tens of thousands of AI data centers equivalent to ~7 trillion RMB ($1 trillion) in saved investment. Critically, this strategy shifts dependency from scarce, expensive GPU/HBM—where China lags—toward more accessible storage, caching, and systems engineering where domestic suppliers like CXMT are gaining strength. Rather than "replacing Nvidia," DeepSeek rebalances AI's value chain away from monolithic hardware dependency. Ultimately, DeepSeek's technical breakthroughs could lower the barrier to AI adoption across Chinese industries by making advanced capabilities affordable at scale—transforming who can access next-generation AI.

marsbit47m ago

Can DeepSeek Save China One Trillion Dollars?

marsbit47m ago

Overturning the Mainstream Approach to Hallucinations: Metacognition is the New Solution for Large Models to Break the Hallucination Barrier

This paper, "Hallucinations Undermine Trust; Metacognition is a Way Forward," proposes a paradigm shift in combating AI hallucination. It argues that the current mainstream approaches—striving for omniscience by scaling data/models or having AI abstain from uncertain answers—are fundamentally flawed. The former has inevitable knowledge gaps, while the latter imposes a crippling "utility tax," requiring the rejection of many correct answers to achieve high accuracy, due to models' poor "discrimination" (the ability to distinguish correct from incorrect answers internally). The core contribution is redefining hallucination not as "being wrong," but as "expressing false information with unwarranted certainty." The proposed solution is **Faithful Uncertainty** or **Metacognition**: enabling AI to accurately perceive its internal uncertainty and honestly express it in its language (e.g., using hedging phrases when unsure). This creates a more reliable assistant that provides useful information while signaling its confidence, minimizing harm from errors. The paper emphasizes that metacognition is critical for the era of AI Agents. Without it, Agents cannot intelligently decide when to use tools like search engines, leading to inefficiency and misuse. Key implementation challenges are highlighted: the "bootstrapping paradox" of training with static uncertainty data, the "alignment distortion signal" where human preference training suppresses internal uncertainty cues, and the difficulty of causally evaluating true metacognition vs. its superficial imitation. The paper concludes that the goal should not be an infallible AI, but one that is honest about the limits of its knowledge, thereby building user trust through transparent communication of its certainty.

marsbit52m ago

Overturning the Mainstream Approach to Hallucinations: Metacognition is the New Solution for Large Models to Break the Hallucination Barrier

marsbit52m ago

Hedge by Buying Gold and Oil, Chase Soaring Returns with AI. ‘Dated’ Bitcoin Enters a Bear Market

Bitcoin has recently declined, hitting a two-month low near $66,123, while Ethereum fell to a three-month low around $1,837. Analysts suggest the drop is not merely due to factors like ETF outflows or MicroStrategy's selling but reflects a deeper issue: Bitcoin is losing a broader asset competition. In a near-zero interest rate environment, Bitcoin previously thrived as an outlet for investor dissatisfaction with inflation and limited options. However, the market landscape has shifted. Bitcoin now occupies an "awkward middle ground," facing competition on three fronts. For inflation hedging, investors prefer gold, energy stocks, and commodity producers—assets with tangible backing and clearer pricing power. For growth exposure, AI-related companies with actual revenues and profits are more attractive. Even within crypto, investors can choose stablecoins, exchanges, or infrastructure firms tied directly to adoption, offering clearer business models and leverage. Thus, Bitcoin is no longer the top choice for hedging, growth, or crypto exposure. This shift is evident in market reactions: despite recent warnings about persistent inflation from a Fed official, Bitcoin did not rally as it might have in the past. Instead, capital flowed to assets with direct commodity or energy exposure. The recent ETF outflows and MicroStrategy sales are symptoms, not causes, of this new reality. Investors are becoming more selective, demanding clearer value propositions beyond mere scarcity. The emerging bear case for Bitcoin is not about it being a bubble or failed technology, but that scarcity alone is no longer sufficient.

华尔街日报54m ago

Hedge by Buying Gold and Oil, Chase Soaring Returns with AI. ‘Dated’ Bitcoin Enters a Bear Market

华尔街日报54m ago

Trading

Spot
Futures

Hot Articles

Discussions

Welcome to the HTX Community. Here, you can stay informed about the latest platform developments and gain access to professional market insights. Users' opinions on the price of AI (AI) are presented below.

活动图片