Some widely-reported milestones.

Each year, AI capabilities that experts predicted were decades away arrive months later. These are widely-reported milestones. Whether they imply what some commentators claim is a separate question.

2017

Transformer architecture invented at Google

The architecture behind every modern AI

Attention Is All You Need - Vaswani et al.
2020

GPT-3 demonstrates emergent reasoning

175B parameters. Few-shot learning surprises researchers.

OpenAI
2022

ChatGPT reaches 100M users in 2 months

Fastest-growing consumer application in history

Reuters
2023

GPT-4 passes bar exam, medical boards, CPA exam

AI outperforms 90th percentile of human professionals

OpenAI Technical Report
2024

Claude, Gemini, and open-source models reach near-expert level

AI coding, writing, and analysis becomes routine

Multiple benchmarks
2025

AI agents begin automating multi-step workflows

AI doesn't just answer - it acts

Anthropic, OpenAI, Google
2027

Projected: AI matches median human performance broadly

Leopold Aschenbrenner's "Situational Awareness" timeline

situational-awareness.ai
PROJECTED
2030

Projected: Artificial General Intelligence

Multiple credible researchers predict AGI by this date

Metaculus, Kurzweil, Hinton
PROJECTED

What some prominent researchers have publicly said.

Paraphrased summaries below, not direct quotes. Views shown are not unanimous across the field - they are a sample of public statements that have been influential in the discourse.

Paraphrased from public statements — not direct quotes. Follow source links to read originals. Read our disclaimers and terms.

Has publicly argued that AI capability is accelerating quickly enough that AI-capable AI research could plausibly arrive within a few years, with significant downstream implications.

Leopold Aschenbrenner
Former OpenAI researcher
Situational Awareness (2024) Read primary source →

Has publicly raised concerns about the long-term safety of systems that may eventually exceed human cognitive capability and argued the topic deserves serious attention.

Geoffrey Hinton
Turing Award winner, often referred to as a "godfather of deep learning"
Public interviews following his 2023 departure from Google Read primary source →

Has long argued for specific predictions about when computers may reach human-level intelligence and what he terms the singularity.

Ray Kurzweil
Author and futurist
The Singularity Is Nearer (2024) Read primary source →

Has publicly suggested that broadly human-capable AI for intellectual work may arrive on a relatively short timeline, while emphasizing safety work.

Dario Amodei
CEO of Anthropic
Public statements (2024) Read primary source →

Has publicly described the emerging period as one of potentially historic technological transformation, with both upside and risk.

Sam Altman
CEO of OpenAI
Public blog posts (2024) Read primary source →

Has publicly emphasized that very powerful AI systems are being built and that safety should be taken seriously.

Ilya Sutskever
Co-founder of OpenAI; founder of Safe Superintelligence Inc.

Recommended Reading

Situational Awareness by Leopold Aschenbrenner has been influential in the discourse around AI timelines. A former OpenAI researcher's argument for why he thinks rapid AI progress is likely. Read it and form your own view - it is one author's argument, not a settled forecast.

~2 hour read. Worth every minute.

Illustrative benchmark figures

AI performance on various benchmarks has improved substantially over recent years. The numbers below are illustrative reference figures, not a settled comparison - benchmarks have known limitations and results vary by methodology.

Task Human AI Year
Legal bar exam 68% 90% 2023
Medical diagnosis (radiology) 87% 94% 2024
Code generation (HumanEval) ~100% 97% 2024
Scientific reasoning (GPQA) 65% 59% 2024
Creative writing (blind test) 46% 54% 2024
Mathematical reasoning (MATH) 90% 96% 2025

Figures shown are illustrative and drawn from public reporting by labs and benchmark authors. Benchmark results vary by version, methodology, prompting, and which "human" baseline is used. Treat all numbers as rough indicators, not authoritative measurements. Verify any figure you intend to rely on against the original source.

So what might one do with this?

One possible response: rather than only competing with AI for cognitive work, consider using AI as a tool while also investing in tangible capabilities. This is the author's framing, not a recommendation.

AUTHOR'S FRAMING

The same technologies changing knowledge work also lower the cost of building, learning, and producing. Whether that matters for any specific person depends on their situation. Not advice.

About this project's book

Note: The paraphrased positions above are from independent third parties and do not constitute endorsements of this project or its book. This section is about the author's own work, separate from the views shown above.

THE BLUEPRINT Surviving the Singularity
A Practical Guide to
Material Independence

What's inside:

  • 01 Economic analysis - the author's reading of recent cost trends
  • 02 The Shouse model - one approach to lowering housing overhead
  • 03 Creator work as a possible income channel (outcomes vary)
  • 04 Local AI tools and the case for running them yourself
  • 05 Land and construction notes (consult your local attorney and contractor)
  • 06 Open-source automation projects worth knowing about
  • 07 Small-product business notes (not a guaranteed income model)
  • 08 A suggested sequence the author has been thinking about

The core blueprint is free to read online. The book is an extended treatment of the same material - informational only, not advice. See full disclaimer.

People are already building.

$0
Cost to start a YouTube channel
Phone + free editing software. That's it.
$5K-$30K
Rural land (1-5 acres)
Multiple counties across 30+ states
$2K
GPU for local AI
Consumer GPUs can run open-weights models locally (capability varies)
114M+
YouTube channels worldwide
The platform is free to publish on; most channels earn little or nothing

Two paths. Same goal. Very different math.

Traditional Path
  • $500K house → $1.1M after interest
  • 30-year commitment
  • Requires steady employment
  • Dependent on employer, market, economy
  • No productive capacity
Total: ~$1.1M+ over 30 years
Builder Path
  • $5-30K land + $25-100K shop/shouse
  • 2-5 year build timeline
  • YouTube funds the build in real-time
  • Own your land, tools, and AI stack
  • Productive from day one
Total: $30K-$150K, own everything

The window is open.
Start building.

You don't need to quit your job tomorrow. You don't need to buy land this week. But you do need to start. Read the blueprint. Run the numbers. See if this path makes sense for you.

The blueprint is free. Always will be. The book is for people who want to go deeper.