Do AI capability breakthroughs arrive on the short timelines people forecast?

Jeff Seibert · Nov 22, 2023

Genuinely useful AI has arrived now, not years away as he believed twelve months ago

The speed of the last generational jump was astonishing — less than a year

Scope: stated as a belief he has revised

54:17 20VC: Why OpenAI Will Become an Infrastructure Play, Why Apple Will Win in an AI World, Why Google is the Most Vulnerable Incumbent, Will LLMs Be Commoditised, Which Startups Are Thin vs Thick Wrappers on Top of LLMs with Jeff Seibert, Founder @ Digits

Kyle Norton · May 30, 2025 · hedged

O3 is potentially AGI, and the AI capability roadmap is moving faster than he anticipated

Rebuilding his ChatGPT workflows on O3 revealed a stark step change in memory, reasoning, speed and ability to research and deeply reason at once

Scope: 'potentially' AGI

72:01 20Sales: From $2M ARR to $40M ARR: The Playbook | How to Use AI To Supercharge Your Sales Team | Why Pipeline Reviews are BS | The Sales Call Script that Closes 99% of Prospects and How to Hire the Best Sales Reps with Kyle Norton, CRO @ Owner

Varun Mohan · Jun 2, 2025 · hedged

People underestimate the speed at which model capability is improving — within six months agents will likely be trusted to write to internal databases and systems at scale

The exponential is hard for people to grasp; only eight or nine months ago people were barely using agents at all

Scope: six-month horizon is approximate

42:35 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

Joelle Pineau · Nov 3, 2025

Despite the perception that AI moves at lightning speed, proving out a fundamental research hypothesis often takes years.

An idea only makes a difference once the right optimizer, compute, and data come together, which takes time to mature.

Scope: about fundamental AI research

3:18 20VC: Cohere's Chief AI Officer on Why Scaling Laws Will Continue | Whether You Can Buy Success in AI with Talent Acquisitions | The Future of Synthetic Data & What It Means for Models | Why AI Coding is Akin to Image Generation in 2015 with Joelle Pineau

Joelle Pineau · Nov 3, 2025

Foundational AI ideas require patience: they can sit dormant for decades until the right algorithmic tweak, context, and problem domain unlock them.

She worked on reinforcement learning for twenty years before reasoning models and agents made everyone talk about it.

4:06 20VC: Cohere's Chief AI Officer on Why Scaling Laws Will Continue | Whether You Can Buy Success in AI with Talent Acquisitions | The Future of Synthetic Data & What It Means for Models | Why AI Coding is Akin to Image Generation in 2015 with Joelle Pineau

Eiso Kant · Oct 7, 2024

Full AGI — machine intelligence surpassing humans across all capabilities — is still quite far out; we will first arrive at human-level capability in a few massively economically valuable domains rather than uniformly

Capabilities are gated by how models learn: they need orders of magnitude more data than humans, so they only become strong where huge datasets exist

Scope: capabilities will not be equally distributed across domains

4:07 20VC: Raising $500M To Compete in the Race for AGI | Will Scaling Laws Continue: Is Access to Compute Everything | Will Nvidia Continue To Dominate | The Biggest Bottlenecks in the Race for AGI with Eiso Kant, CTO @ Poolside

Jonathan Siddharth · Dec 1, 2025

We are in innings one of verticalized data acquisition, and AGI and superintelligence will arrive via slow, steady takeoff rather than rapid takeoff.

It will take a long time to get all of this workflow data into the models.

11:40 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

Harry Stebbings · Nov 3, 2025

Leading AI figures are now tempering their timelines — Karpathy calling it the decade of agents rather than the year of agents, and Altman similarly pulling back.

3:47 Even the optimists are now lengthening their timelines

Tom Hulme · May 8, 2024 · hedged

There is a correlation between how aggressively someone predicts AGI is imminent and how much capital they need to raise, so you should weight the views of those who don't need to raise money

Amodei, Altman and Musk all need capital and say AGI is around the corner, while Zuckerberg and Hassabis, who don't need to raise, are far more cautious — assuming equal intelligence, the non-fundraisers' views are less distorted

68:27 Capital raising need correlates with aggressive agi timeline claims

David Cahn · Oct 27, 2025

The people who invented the AI paradigm hold the longest timelines while the newest entrants hold the shortest, because lab culture confers status on whoever gives the most aggressive AGI date

There's an experience curve — people who have done this longest know these things are hard and take time

33:56 Veteran ai researchers hold longer timelines newcomers chase status with aggressive dates

Ethan Mollick · Jul 31, 2024

Stated AGI timelines should be taken with a grain of salt because people are highly self-motivated; the credible signal is people staking their careers and reputations on it being possible

People are very self-motivated in what they claim, but reputational and career bets are costly and therefore informative — and historically AGI has always been said to be soon, going back to the 1957 Dartmouth conference

41:13 Career and reputation stakes not mere statements signal genuine agi belief

Richard Socher · Apr 18, 2025 · hedged

He will probably still win his bet that AGI-level milestones — an affordable house-cleaning personal robot, solving a millennium math problem, and a book translation the author would endorse as official — will not all be met by 2027

He was skeptical of the earlier 'steps towards AGI' framing of robotic hands and DOTA game-playing, seeing them as cool robotics and RL projects rather than AGI progress

51:57 Specific milestone tests show 2027 agi bets will fail

Victor Lazarte · Apr 14, 2025

AI is going to happen far faster than people believe, with lives meaningfully changed within five years

Three years ago ChatGPT didn't exist and now close to a billion people use it monthly with a meaningfully different experience

42:01 Ai changes daily life meaningfully within five years

Alex Schultz · Sep 5, 2025

AI progress is predictable roughly two innovations ahead, just as it was in semiconductors, and beyond that horizon nobody can forecast which lab techniques will break through

In chip history each generation could see about two steps ahead; agentic behavior and AI engineers were being discussed 1.5-2 years before they arrived, and even OpenAI didn't know GPT-2 would be the breakthrough it was on launch day

13:24 Ai progress is predictable about two innovations ahead like semiconductor history