Kazuki
@kazuki
Cofounder of Glasp. I collect ideas and stories worth sharing 📚
San Francisco, CA
Joined Oct 9, 2020
1072
Following
5817
Followers
1.47k
13.57k
171.17k
Atomic Graph
New AI classifier for indicating AI-written text
openai.com/blog/new-ai-classifier-for-indicating-ai-written-text/
Jan 31, 2023
6
How I’m Letting Internet Strangers Find Smarter Articles For Me via Glasp
benjaminboman.medium.com/how-im-letting-internet-strangers-find-smarter-articles-for-me-via-glasp-c20bda710bff
Jan 31, 2023
61
Startup Priorities
blog.geoffralston.com/startup-priorities
Jan 28, 2023
6
Prompt-Driven Design | News | Felicis Ventures
www.felicis.com/news/prompt-driven-design
Jan 27, 2023
7
Advice for Building in AI
every.to/divinations/advice-for-building-in-ai
Jan 25, 2023
7
Who Owns the Generative AI Platform? | Andreessen Horowitz
a16z.com/2023/01/19/who-owns-the-generative-ai-platform/
Jan 22, 2023
134
For a More Creative Brain, Follow These 5 Steps
jamesclear.com/five-step-creative-process
Jan 20, 2023
92
Enterprise Software Is Dead. Long Live Enterprise Software.
digitalnative.substack.com/p/enterprise-software-is-dead-long
Jan 19, 2023
91
AI in 2023: The Application Layer Has Arrived
digitalnative.substack.com/p/ai-in-2023-the-application-layer
Jan 18, 2023
91
Why the Best Things in Life Are All Backwards
markmanson.net/the-backwards-law
Jan 11, 2023
133
Personal Viewpoint - CM #2
curationmonetized.substack.com/p/personal-viewpoint-cm-2
Jan 10, 2023
1
Lessons From The Samurai: The Secret To Always Being At Your Best - Barking Up The Wrong Tree
bakadesuyo.com/2014/02/samurai/
Jan 9, 2023
81
The End of Organizing
every.to/superorganizers/the-end-of-organizing
Jan 8, 2023
102
Labeling Yourself is Keeping You Down, Do This Instead
www.nirandfar.com/labeling-yourself
Jan 6, 2023
63
Predicting machine learning moats
robotic.substack.com/p/ml-moats
Dec 30, 2022
9
How to Improve Your Creative Thinking
every.to/superorganizers/understanding-the-science-of-creativity
Dec 26, 2022
162
This Artificial Intelligence (AI) Application Does YouTube Summary with ChatGPT
www.marktechpost.com/2022/12/22/this-artificial-intelligence-ai-application-does-youtube-summary-with-chatgpt/
Dec 23, 2022
5
We Evaluated ChatGPT vs. Google on 500 Search Queries
www.surgehq.ai/blog/googles-existential-threat-chatgpt-matches-googles-performance-on-informational-search-queries-and-smashes-it-on-coding
Dec 21, 2022
6
YouTube Summaries | Ben's Bites
www.bensbites.co/p/youtube-summaries
Dec 19, 2022
11
Generative Tech Market Map and 5-Layer Tech Stack
www.nfx.com/post/generative-ai-tech-5-layers
Dec 19, 2022
192
Everything is Aiming
nesslabs.com/everything-is-aiming
Dec 17, 2022
61
Introducing Text and Code Embeddings
openai.com/blog/introducing-text-and-code-embeddings/
Dec 17, 2022
51
ChatGPT and the Imagenet moment — Benedict Evans
www.ben-evans.com/benedictevans/2022/12/14/ChatGPT-imagenet
Dec 17, 2022
4
Glasp: A Web Clipper & Web Highlighter Makes My Life Easier -
deskoflawyer.com/glasp-web-clipper-web-highlighter/
Dec 12, 2022
8
Overview & Applications of Large Language Models (LLMs)
leighmariebraswell.substack.com/p/overview-and-applications-of-large
Dec 12, 2022
71
The brief history of artificial intelligence: The world has changed fast – what might be next?
ourworldindata.org/brief-history-of-ai
Dec 10, 2022
61
The Day The AGI Was Born
lspace.swyx.io/p/everything-we-know-about-chatgpt
Dec 6, 2022
62
ChatGPT Is the World’s Best Chatbot
thealgorithmicbridge.substack.com/p/chatgpt-is-the-worlds-best-chatbot
Dec 5, 2022
31
ChatGPT: Optimizing Language Models for Dialogue
openai.com/blog/chatgpt/
Nov 30, 2022
72
An OpenAI update | Ben's Bites
bensbites.beehiiv.com/p/openai-update
Nov 29, 2022
11
Malcolm Gladwell’s Top 13 Writing Tips
writingcooperative.com/malcolm-gladwells-top-13-writing-tips-c833432c005f
Nov 29, 2022
164
The first AI to laugh and Lex Fridman interviews Richard Feynman | Ben's Bites
bensbites.beehiiv.com/p/first-ai-laugh-lex-fridman-interviews-richard-feynman
Nov 25, 2022
11
The psychology of prestige: why we play the social status game
nesslabs.com/prestige-psychology
Nov 24, 2022
91
6 New Theories About AI
every.to/napkin-math/6-new-theories-about-ai
Nov 24, 2022
142
Catching Unicorns with GLTR
gltr.io/
Nov 23, 2022
61
Generative Tech Begins
www.nfx.com/post/generative-tech
Nov 22, 2022
81
Make Highlights And Annotations Social With Glasp The Social Highlighter
medium.com/taking-notes/make-highlights-and-annotations-social-with-glasp-the-social-highlighter-6028ff0358c2
Nov 22, 2022
4
Building for Believers
davidspinks.substack.com/p/building-for-believers
Nov 18, 2022
143
The New Learning Economy: It’s Time To Build in Education | Andreessen Horowitz
a16z.com/2022/09/16/the-new-learning-economy-its-time-to-build-in-education/
Nov 18, 2022
101
Generative Tech Market Map and 5-Layer Tech Stack
Highlights & Notes
- Applications layer
- OS or API layer
- Hyperlocal AI models
- Specific AI models
- General AI models
General AI models are the core technology breakthrough. It’s something like GPT-3 for text, DALL-E-2 for images, Whisper for voice, or Stable Diffusion. These models deal with broad categories of outputs: text, images, videos, speech, games.
Specific AI models capture even more nuance for specific jobs such as writing tweets, ad copy, song lyrics, or generating e-commerce photos, 3D interior design images, etc. These models are trained on more narrow, more specialized data
Hyperlocal AI models are specialists. A hyperlocal AI model can write a scientific article in the style preferred by Nature. It creates interior design models suited to a specific person’s aesthetic. It can write code in the particular style of an individual company.
data does not always provide a powerful defensibility. Even if a competitor cannot get your exact dataset, they likely can find a similar dataset. Even if their model is not quite as good as yours, customers can’t always tell, and competitors can claim to have what you have in their sales materials.
most data network effects asymptote over time. An AI model that is 5% or even 20% percent better than the competition is a pretty slim defensibility.
People’s brains won’t be able to tell the difference between human writing and AI writing within 24 months. Most people will enjoy music and lyrics written by AI in 36 months.
the best place to explore data network effects in your AI models is probably at level 3, the hyperlocal layer, which benefits from proprietary and trusted data.
API layer or Generative OS helps that application access all the AI models the application needs. This layer also allows for the AI models to be switched out at will. Which of course tends to commodify them.
There will be 10,000’s of these applications built for various needs in the next 2 years. Incumbent software providers will add generative features. New companies will create competitors to the old, emphasizing generative as a wedge.
- Product Speed
You need to get product in market, see what works, what doesn’t. See what makes people uncomfortable – and get them through that cycle. Watch your competitors closely and borrow the best ideas. Don’t make it perfect. Don’t spend too much time hunting down specific data in hopes of building the perfect model
Launch the feature first, let the model learn over time.
- Fundraising Speed
- Sales Speed
Aggressive sales will help embed your product in your customers and give you the right to expand into other categories. Aggressive sales will help you build network effects to help your defensibility. Sales will help you with advantages #2 and #1 above.
- Network Effects
- Embedding
- Find an investor who will sprint with you