andrewthompsonx's profile picture. CTO of @orbital_news 🇬🇧 (Vertical AI Agent for real estate legal). Advises startups & VCs. Founded an AI company. Ships early & often. 🚲 ☕️ 🔧 ⏱ 🚀

Andrew Thompson

@andrewthompsonx

CTO of @orbital_news 🇬🇧 (Vertical AI Agent for real estate legal). Advises startups & VCs. Founded an AI company. Ships early & often. 🚲 ☕️ 🔧 ⏱ 🚀

Andrew Thompson reposted

smooth exponential AI progress

Things are looking smoothly exponential for AI over the past several years, and I continue to think this is the best default assumption (until the AI R&D automation feedback loop eventually speeds everything up)

daniel_271828's tweet image. Things are looking smoothly exponential for AI over the past several years, and I continue to think this is the best default assumption (until the AI R&D automation feedback loop eventually speeds everything up)


Andrew Thompson reposted

What makes a great company name? It’s a question I spend more time thinking about than most people. Here are some of my favorites:

hiiamArielle's tweet image. What makes a great company name? It’s a question I spend more time thinking about than most people. Here are some of my favorites:

Andrew Thompson reposted

great chart from @EpochAIResearch via @a16z. the "Moore's Law of AI" is accelerating - from going 9x cheaper per year for GPT3.5 level models, to 40x cheaper per year for GPT4 level, to 900x cheaper per year for 4o level models. did a quick check and think o-series reasoning…

swyx's tweet image. great chart from @EpochAIResearch via @a16z. 

the "Moore's Law of AI" is accelerating - from going 9x cheaper per year for GPT3.5 level models, to 40x cheaper per year for GPT4 level, to 900x cheaper per year for 4o level models.

did a quick check and think o-series reasoning…
swyx's tweet image. great chart from @EpochAIResearch via @a16z. 

the "Moore's Law of AI" is accelerating - from going 9x cheaper per year for GPT3.5 level models, to 40x cheaper per year for GPT4 level, to 900x cheaper per year for 4o level models.

did a quick check and think o-series reasoning…

Andrew Thompson reposted

Every product leader should have a "Things I Believe" page like @leerob. I think I'm going to write one but it might be too similar to his 😅

petergyang's tweet image. Every product leader should have a "Things I Believe" page like @leerob. 

I think I'm going to write one but it might be too similar to his 😅

.@sama and @satyanadella on BG2 podcast on the decreasing cost of compute youtu.be/Gnl833wXRz0?si…

andrewthompsonx's tweet image. .@sama and @satyanadella on BG2 podcast on the decreasing cost of compute

youtu.be/Gnl833wXRz0?si…

Andrew Thompson reposted

Government surveys say 9% of US businesses use AI: @tryramp data says it's closer to 44%. Based on millions of aggregated + anonymized card + AP transactions, invoices, receipts, POs, we're seeing real adoption, especially in construction firms, manufacturers, and health card…

iandmacomber's tweet image. Government surveys say 9% of US businesses use AI: @tryramp data says it's closer to 44%. Based on millions of aggregated + anonymized card + AP transactions, invoices, receipts, POs, we're seeing real adoption, especially in construction firms, manufacturers, and health card…

Andrew Thompson reposted

It looks like AI music is following the same path as AI text: 1) Appears to have passed the Turing Test, people are only 50/50 in identifying older Suno vs. human songs (but 60/40 when two songs are the same genre) 2) Same fast development, new models are getting better quickly.

emollick's tweet image. It looks like AI music is following the same path as AI text:
1) Appears to have passed the Turing Test, people are only 50/50 in identifying older Suno vs. human songs (but 60/40 when two songs are the same genre)
2) Same fast development, new models are getting better quickly.
emollick's tweet image. It looks like AI music is following the same path as AI text:
1) Appears to have passed the Turing Test, people are only 50/50 in identifying older Suno vs. human songs (but 60/40 when two songs are the same genre)
2) Same fast development, new models are getting better quickly.
emollick's tweet image. It looks like AI music is following the same path as AI text:
1) Appears to have passed the Turing Test, people are only 50/50 in identifying older Suno vs. human songs (but 60/40 when two songs are the same genre)
2) Same fast development, new models are getting better quickly.

Andrew Thompson reposted

The most important skill in the 21st century is the ability to live happily with uncertainty.


Andrew Thompson reposted

I quite like the new DeepSeek-OCR paper. It's a good OCR model (maybe a bit worse than dots), and yes data collection etc., but anyway it doesn't matter. The more interesting part for me (esp as a computer vision at heart who is temporarily masquerading as a natural language…

🚀 DeepSeek-OCR — the new frontier of OCR from @deepseek_ai , exploring optical context compression for LLMs, is running blazingly fast on vLLM ⚡ (~2500 tokens/s on A100-40G) — powered by vllm==0.8.5 for day-0 model support. 🧠 Compresses visual contexts up to 20× while keeping…

vllm_project's tweet image. 🚀 DeepSeek-OCR — the new frontier of OCR from @deepseek_ai , exploring optical context compression for LLMs, is running blazingly fast on vLLM ⚡ (~2500 tokens/s on A100-40G) — powered by vllm==0.8.5 for day-0 model support.

🧠 Compresses visual contexts up to 20× while keeping…
vllm_project's tweet image. 🚀 DeepSeek-OCR — the new frontier of OCR from @deepseek_ai , exploring optical context compression for LLMs, is running blazingly fast on vLLM ⚡ (~2500 tokens/s on A100-40G) — powered by vllm==0.8.5 for day-0 model support.

🧠 Compresses visual contexts up to 20× while keeping…
vllm_project's tweet image. 🚀 DeepSeek-OCR — the new frontier of OCR from @deepseek_ai , exploring optical context compression for LLMs, is running blazingly fast on vLLM ⚡ (~2500 tokens/s on A100-40G) — powered by vllm==0.8.5 for day-0 model support.

🧠 Compresses visual contexts up to 20× while keeping…


Andrew Thompson reposted

The graveyard is littered with AI products that got copied and lost their users without building moats. That's why I invited Miqdad (OpenAI product leader) to share his 7-step playbook to build defensible AI products: 1. Define your arena 2. Design the user loop (not the AI…

petergyang's tweet image. The graveyard is littered with AI products that got copied and lost their users without building moats.

That's why I invited Miqdad (OpenAI product leader) to share his 7-step playbook to build defensible AI products:

1. Define your arena
2. Design the user loop (not the AI…

Andrew Thompson reposted

Everyone should be using Claude Code more PMs, marketers, designers, founders, parents. Everyone. The trick is to forget that it’s called Claude Code and instead think of it as Claude Local or Claude Agent. It’s essentially a super-intelligent AI running locally, able to do…

lennysan's tweet image. Everyone should be using Claude Code more

PMs, marketers, designers, founders, parents. Everyone.

The trick is to forget that it’s called Claude Code and instead think of it as Claude Local or Claude Agent. It’s essentially a super-intelligent AI running locally, able to do…

Andrew Thompson reposted

These are the top 15 most valued private AI companies, their latest public revenue and growth.

deedydas's tweet image. These are the top 15 most valued private AI companies, their latest public revenue and growth.

Andrew Thompson reposted

The next billion dollar agents will be what I call “proactive agents.” These are the ones capable of predicting user intent before a request. No one has figured out a reliable way to do this just yet. We are close.


Andrew Thompson reposted

Hearing from a lot of good founders that AI tools are writing most of their code now. Software engineers orchestrate the AI. They are also finding it extremely hard to hire because most experienced engineers have their heads in the sand and refuse to learn the latest tools.


.@dwarkesh_sp on the future of AI and when we will see intelligence explosion youtu.be/VwLE2KqX9xU?si…

andrewthompsonx's tweet image. .@dwarkesh_sp on the future of AI and when we will see intelligence explosion youtu.be/VwLE2KqX9xU?si…

Andrew Thompson reposted

OpenAI just dropped the top 30 customers who've used 1T+ of their tokens.

deedydas's tweet image. OpenAI just dropped the top 30 customers who've used 1T+ of their tokens.

Andrew Thompson reposted

It can be hard to viscerally convey the speed of AI progress.This video does a great job! 2 years ago: LLMs could barely code. Last year: they got pretty good at it. This year: they can build our actual product. Everyone needs to internalize this pace.

We asked every version of Claude to make a clone of Claude(dot)ai, including today’s Sonnet 4.5… see what happened in the video



Loading...

Something went wrong.


Something went wrong.