Is the binding constraint on AI usefulness model capability or the products built around the models?

Pat Grady · Jul 8, 2024

Even if foundation model capabilities were frozen today, optimization, tuning and developer ecosystem work alone would revolutionize trillions of dollars of industries and create trillions to tens of trillions of dollars of market cap

Current model capabilities are already powerful beyond anything people have figured out how to do with them

Scope: hypothetical freeze of capabilities; the capability race is in fact still ongoing; offered as a view the data does not yet support

63:25 20VC: The Sequoia Investment Process | Investing Lessons from Doug Leone, Roelof Botha & Alfred Lin | Sequoia's Framework for Analysing Founders | The True Benefit of Having Sequoia on a Cap Table & Sequoia's Biggest Threat with Pat Grady

Varun Mohan · Jun 2, 2025

Model availability alone doesn't create the product — Windsurf's agentic IDE required significant work on its own models and UI/UX before the experience took off

Models like Sonnet had existed for a while, yet it still took work to combine them with internal models and review UX to make the experience usable

46:32 20VC: Windsurf Founder on Will Model Companies Own the App Layer | Why Moats Do Not Exist in a World of AI | Why the Notion of Single Person $BN Companies is BS | Lovable vs Bolt & Cursor vs Windsurf: How Does it All End with Varun Mohan

Max Junestrand · Aug 15, 2025

The product-to-model gap is massive — products use less than 20% of what today's models can do, so there are years of building left even with zero further model improvement

Applications simply haven't caught up to existing model capability

Scope: estimate of under 20% utilization

8:32 20VC: 15 Term Sheets in 7 Days and Choosing Benchmark | Harvey vs Legora: Who Wins Legal and How to Play When You Have $600M Less Funding | Are AI Models Plateauing Today | Building a 9-9-6 Culture From Stockholm with Max Junestrand

Max Junestrand · Aug 15, 2025

Products will increasingly use models in far more creative and parallelized ways than today's basic single-call usage

Current usage is still quite basic; parallelizing tasks like due diligence across documents and questions already works

10:31 20VC: 15 Term Sheets in 7 Days and Choosing Benchmark | Harvey vs Legora: Who Wins Legal and How to Play When You Have $600M Less Funding | Are AI Models Plateauing Today | Building a 9-9-6 Culture From Stockholm with Max Junestrand

Maor Shlomo · Nov 24, 2025

Even with today's models we are only scratching the surface of the economic value that can be created

A one-person team could be sold for $80M without people, so much more can be done with far fewer people

0:00 20VC: Base44's Maor Shlomo on How Vibe Coding Will Kill SaaS and Salesforce | Why it is BS that Vibe Coding Platforms Do Not Have Defensibility and Bad Margins | Why He Worries About Google, Not Replit and Lovable | Why Long Anthropic, Not OpenAI?

Jonathan Siddharth · Dec 1, 2025

There is a significant model capability overhang: models can do X but users only extract X minus delta, and the right agentic scaffold unlocks the gap without any new model

Correct system prompts, user prompts, context access, context acquisition and internal tool use dramatically expand what today's models deliver — e.g. a fine-tuned agent could do the clip-selection work Harry does manually

40:00 20VC: Scale, Surge, Turing, Mercor: Who Wins & Who Loses in Data Labelling | Is Revenue in Data Labelling Real or GMV? | Why 99% of Knowledge Work Will Go and What Happens Then? | Why SaaS is Dead in a World of AI with Jonathan Siddharth @ Turing

Winston Weinberg · Jan 19, 2026

Even if OpenAI and Anthropic stopped shipping new models entirely, their revenues would still explode because the capability overhang of current models is astronomical

People and especially businesses don't know how to use these systems yet; current model capability is far from being integrated into daily life, and many companies will build on top of existing models

Scope: capability overhang is higher than anyone is talking about

25:18 20VC: How Model Performance is Plateauing | Two Key Rules for Effective Deal-Making | Company Building Lessons from Keith Rabois, Brian Halligan and Pat Grady | Why Enterprise AI Adoption is Years Off with Harvey CEO Winston Weinberg

Alexander Embiricos · Feb 21, 2026

There is a large capability overhang in AI models — even models from a year and a half ago were far more useful than the products built around them allowed

Products hadn't been built to unlock the models, so technical people captured most of the benefit while non-technical people like his grandma got little

59:06 20VC: Codex vs Claude Code vs Cursor: Who Wins, Who Loses | Will All Coding Be Automated - Do We Need PMs | The Real Bottleneck to AGI | The Three Phases of Agents and What You Need to Know with Alex Embiricos, Head of Codex at OpenAI