Sign in

Gradient Flow

@gradientflow.com
340 followers 108 following 1.5K posts

Newsletter: gradientflow.substack.com Podcast: thedataexchange.media Blog: gradientflow.com/blog

PostsRepliesMedia
Gradient Flow @gradientflow.com · 9h
What changes when AI becomes a game mechanic instead of a cost-cutting tool? Studio Atelico CEO Piero Molino explores on-device models, post-training, artist royalties, coding agents, and a new approach to AI-native games. thedataexchange.media/piero-molino...
thedataexchange.media
Why AI Makes Games Worse—and How to Make Them Better
Piero Molino on Player Backlash, Small Open-Weight Models, and the Economics of AI in Video Games.
000
Gradient Flow @gradientflow.com · 30/09/2026
The data problem didn't disappear, it moved 🎯 gradientflow.substack.com/p/your-model...
gradientflow.substack.com
Something changed in how AI gets fed
Put data, machine learning, and AI to work.
001
Gradient Flow @gradientflow.com · 30/09/2026
AI's data problem is changing. The raw material may already exist. The harder job is turning it into information that agents and robots can safely and effectively use. 🎯 gradientflow.substack.com/p/your-model...
gradientflow.substack.com
Something changed in how AI gets fed
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 30/09/2026
🎗️ AI's next data problem is less about finding information and more about making it usable. 🔎 Digital agents need trustworthy, permitted inputs, robots need human experience translated into forms they can learn from. gradientflow.substack.com/p/your-model...
gradientflow.substack.com
Something changed in how AI gets fed
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 29/09/2026
🎗️ Who Owns the Work of Making AI Data Safe? 🔎 The AI data problem has shifted. It's no longer about finding raw material. It's about doing the work to make it usable. That's true for robots learning from video and for agents pulling from the web. gradientflow.substack.com/p/your-model...
gradientflow.substack.com
Something changed in how AI gets fed
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 28/09/2026
🚨🚨🚨 the New Yorker on the current state of the US DOJ 🚨🚨🚨 archive.ph/IrbQ7
archive.ph
000
Gradient Flow @gradientflow.com · 28/09/2026
🎗️Six Reasons I Think Open AI Models Will Win 🏆 gradientflow.com/six-reasons-...
gradientflow.com
Six Reasons I Think Open AI Models Will Win - Gradient Flow
In my conversations with developers and AI teams, I’m struck by how many are exploring moving more of their inference workloads to open models. I’ve touched on some of their reasons before , but here ...
011
Gradient Flow @gradientflow.com · 28/09/2026
The story behind Lectric e-bikes 🚲🚲🚲🔋 “We see Lectric as a mobility platform” youtu.be/coyHcx1An_s?...
youtu.be
The Strategy Behind Lectric’s eBike Success | Quantum Marketing
YouTube video by Bloomberg Originals
010
Gradient Flow @gradientflow.com · 27/09/2026
Super impressed with Tesla phone support for our residential powerwall battery for our solar system. They take time to diagnose the problem remotely, and patiently walk you through the steps to fix your system
000
Gradient Flow @gradientflow.com · 27/09/2026
🆕 Post-Transformer AI: Reasoning Without Language Pathway's CEO Zuzanna Stamirowska on why transformers aren't enough for true reasoning. Their BDH model reasons in latent space, learns from experience, maintains state over long horizons - all at 150M parameters thedataexchange.media/zuzanna-stam...
thedataexchange.media
What Comes After Transformers?
Zuzanna Stamirowska on Latent Reasoning, Continual Learning, Small Reasoning Models, and Long-Horizon Agents.
000
Gradient Flow @gradientflow.com · 27/09/2026
🆕 Stale Data Was an Analytics Problem. Agents Make It an Operations Problem A human analyst can notice when a number looks wrong and ask a question. An agent just acts. That changes what data quality actually costs you. gradientflow.substack.com/p/your-ai-mi...
gradientflow.substack.com
why your data stack is about to get less forgiving
Put data, machine learning, and AI to work.
010
Gradient Flow @gradientflow.com · 26/09/2026
Master rhyme master 🔥Rick Glassman prompts Harry Mack #hiphopninja youtu.be/DJWmotvuXBA?...
youtu.be
Girls at the Guggenheim | Flow State with Harry Mack x Rick Glassman
YouTube video by Harry Mack Clips
000
Gradient Flow @gradientflow.com · 25/09/2026
Your AI might be fine but your data is lying to it 🔎 Some AI failures are really old data failures. The AI didn't create the duplicate record or the inconsistent definition. It just acted on them. gradientflow.substack.com/p/your-ai-mi...
gradientflow.substack.com
why your data stack is about to get less forgiving
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 24/09/2026
What Critics Get Wrong About Data Centers 🔎 gradientflow.com/ai-data-cent...
001
Gradient Flow @gradientflow.com · 24/09/2026
THE GOSPEL OF HIGH YIELDS? 🏈 Tim Tebow making up to $100K PER TALK pushing high-pressure investment courses www.youtube.com/watch?v=uh7s...
youtube.com
We Investigated the Christian "Scam" Fronted by Tim Tebow | PTFO
YouTube video by PABLO TORRE FINDS OUT
000
Gradient Flow @gradientflow.com · 24/09/2026
🆕 Post-Transformer AI: Reasoning Without Language Pathway's CEO Zuzanna Stamirowska on why transformers aren't enough for true reasoning. Their BDH model reasons in latent space, learns from experience, and maintains state over long horizons thedataexchange.media/zuzanna-stam...
thedataexchange.media
What Comes After Transformers?
Zuzanna Stamirowska on Latent Reasoning, Continual Learning, Small Reasoning Models, and Long-Horizon Agents.
010
Gradient Flow @gradientflow.com · 24/09/2026
🚨 agents broke a few assumptions about your data stack 🎯 gradientflow.substack.com/p/your-ai-mi...
gradientflow.substack.com
why your data stack is about to get less forgiving
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 23/09/2026
You're using LLMs for the wrong job Learn more 🎯 gradientflow.com/jev-unpacked/
000
Gradient Flow @gradientflow.com · 23/09/2026
AI Agents Raise the Stakes for Data Infrastructure 🎯 A working agent demo can hide how much still needs to be built. Agents don’t just need relevant data. They need to know what is true right now. Did the payment clear? Is the inventory still there? 👉 gradientflow.substack.com/p/your-ai-mi...
gradientflow.substack.com
why your data stack is about to get less forgiving
Put data, machine learning, and AI to work.
010
Gradient Flow @gradientflow.com · 23/09/2026
Enterprise AI has focused heavily on finding relevant information. Agents create an additional requirement: they need information that reflects what is true now. Payment status, inventory, account state, and recent changes can all affect the next action. gradientflow.substack.com/p/your-ai-mi...
gradientflow.substack.com
why your data stack is about to get less forgiving
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 22/09/2026
🆕 Stale Data Was an Analytics Problem. Agents Make It an Operations Problem A human analyst can notice when a number looks wrong and ask a question. An agent just acts. That changes what data quality actually costs you. gradientflow.substack.com/p/your-ai-mi...
gradientflow.substack.com
why your data stack is about to get less forgiving
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 22/09/2026
In SF's terrain, the crane's campaign: stay level, stay sane
000
Gradient Flow @gradientflow.com · 20/09/2026
Inside OpenAI’s Nine-Month Chip Sprint 🖥️ Richard Ho on designing custom AI hardware for high throughput and low latency thedataexchange.media/richard-ho-o...
thedataexchange.media
Inside OpenAI’s First Chip
Richard Ho on Custom AI Chips, Inference Economics, CUDA, and AI-Assisted Engineering.
000
Gradient Flow @gradientflow.com · 19/09/2026
What Operating Experience Can Teach AI 🔎 Two companies, same AI model. One connects it to documents. The other has thousands of examples of how experienced people do the work, tied to outcomes. That gap compounds over time. gradientflow.substack.com/p/what-count...
gradientflow.substack.com
so Google just paid $10 million for an airline's Teams messages
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 18/09/2026
Mountain Lion captured 🪤 #SanFrancisco www.youtube.com/watch?v=YdpB...
youtube.com
Mountain lion captured in San Francisco's Outer Sunset
YouTube video by NBC Bay Area
010
Gradient Flow @gradientflow.com · 18/09/2026
An Impressive Engineering Feat: Richard Ho, OpenAI’s VP of Hardware, on building the company’s first custom AI chip in just nine months 🎧🎙️ thedataexchange.media/richard-ho-o...
thedataexchange.media
Inside OpenAI’s First Chip
Richard Ho on Custom AI Chips, Inference Economics, CUDA, and AI-Assisted Engineering.
000
Gradient Flow @gradientflow.com · 18/09/2026
Such a moving video. Sad and tragic, but really worth watching. youtu.be/Kimur4jkJT8?... 
youtu.be
The $15,000 Settlement That Destroyed My Father's Life | NYT Opinion
YouTube video by New York Times Opinion
000
Gradient Flow @gradientflow.com · 17/09/2026
Low latency, high throughput, and a 9-month timeline. Watch OpenAI’s VP of Hardware, Richard Ho, unpack the engineering decisions behind their very impressive, first custom silicon. thedataexchange.media/richard-ho-o...
thedataexchange.media
Inside OpenAI’s First Chip
Richard Ho on Custom AI Chips, Inference Economics, CUDA, and AI-Assisted Engineering.
010
Gradient Flow @gradientflow.com · 17/09/2026
State of the Trump/Republican economy at a glance gasprices.aaa.com
000
Gradient Flow @gradientflow.com · 17/09/2026
From Logs to Learning: How Companies Can Turn Process Into an AI Advantage ❝If I were running an enterprise AI program, I would start by identifying work that is valuable, comes up often, requires real expertise, and produces outcomes I can evaluate. gradientflow.substack.com/i/212075983/...
gradientflow.substack.com
so Google just paid $10 million for an airline's Teams messages
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 16/09/2026
🆕 Your Company's Most Valuable AI Asset Might Be How Work Gets Done 🔎 Two companies, same AI model. One connects it to documents. The other has thousands of examples of how experienced people do the work, tied to outcomes. That gap compounds over time. gradientflow.substack.com/p/what-count...
gradientflow.substack.com
so Google just paid $10 million for an airline's Teams messages
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 16/09/2026
Re: Iran War The situation continues to get worse 📈 fuel prices 💰💸 costing $40B and counting youtu.be/l0XDwKBkSzw?...
youtu.be
Iran war sparks global fuel riots; Hegseth faces GOP impeachment push
YouTube video by CNN
000
Gradient Flow @gradientflow.com · 16/09/2026
General Admission is almost sold out, see you at the 4th AI Conference in a few weeks 🙌 aiconference.com
001
Gradient Flow @gradientflow.com · 16/09/2026
What counts as valuable company data is changing 👉 gradientflow.substack.com/p/what-count...
gradientflow.substack.com
so Google just paid $10 million for an airline's Teams messages
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 15/09/2026
Dolores Park, Mission District, San Francisco
000
Gradient Flow @gradientflow.com · 15/09/2026
🆕 Your Company's Most Valuable AI Asset Might Be How Work Gets Done 🔎 Two companies, same AI model. One connects it to documents. The other has thousands of examples of how experienced people do the work, tied to outcomes. That gap compounds over time. gradientflow.substack.com/p/what-count...
gradientflow.substack.com
so Google just paid $10 million for an airline's Teams messages
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 15/09/2026
So cool youtu.be/wAXf-8beX6E?...
youtu.be
How London’s Rarest Crafts Are Still Made Today | Made In | Condé Nast Traveler
YouTube video by Condé Nast Traveler
000
Gradient Flow @gradientflow.com · 15/09/2026
👀 AI Policy Is Now a Hostage of the Epstein Files 🙌 Massie predicted Johnson would recess the House to dodge the Epstein vote. Every pending AI bill died with the calendar. Whatever your AI politics, Congress should at least be arguing about this stuff www.publicnotice.co/p/mike-johns...
publicnotice.co
Mike Johnson runs out the clock on his own speakership
It was the most sycophantic of times.
100
Gradient Flow @gradientflow.com · 13/09/2026
Much of the debate around OpenAI and Navier-Stokes has focused on the math community’s reaction. I think the bigger story is what comes next for knowledge work. If AI can produce the output, how do we decide who actually understands the work? thedataexchange.media/quick-take-2...
thedataexchange.media
When Everyone Is 10x More Productive, No One Is
Ben Lorica and Evangelos Simoudis on AI IP Protection, Navier-Stokes Controversy, Human-Reserved Roles, and Workplace Involution.
000
Gradient Flow @gradientflow.com · 13/09/2026
Recursive self-improvement gets attention, but application teams may see useful forms of automated improvement much sooner. 🙌 gradientflow.substack.com/p/this-is-ho...
gradientflow.substack.com
Your AI model is a rental (but the loop is an asset)
Put data, machine learning, and AI to work.
000
Gradient Flow @gradientflow.com · 13/09/2026
This is so strange #germany youtu.be/vgJ8kyttaQs?...
youtu.be
What Germany gets wrong about Native Americans
YouTube video by DW Euromaxx
000
Gradient Flow @gradientflow.com · 12/09/2026
Cars are becoming giant touchscreens. But on-road studies find infotainment systems can impose high visual and cognitive demands. The risk depends as much on interface design as screen size. A useful look at where car design may be heading: mwg.aaa.com/via/car/info...
mwg.aaa.com
Are Big Infotainment Screens Here to Stay? | Via
Displays in cars are hitting record sizes, while tactile controls like buttons are vanishing. Are large screens the future or are buttons going to make a comeback?
000
Gradient Flow @gradientflow.com · 12/09/2026
Self-Improving AI: What's Actually Useful Right Now 🎯 Treat every automated AI prompt or workflow tweak like a standard software deployment. Run independent tests, maintain rollback options, and require human sign-off on major changes. gradientflow.substack.com/p/this-is-ho...
gradientflow.substack.com
Your AI model is a rental (but the loop is an asset)
Put data, machine learning, and AI to work.
010
Gradient Flow @gradientflow.com · 12/09/2026
The most interesting thing about the Navier-Stokes story may not be the math. It may be what it previews for knowledge work: when AI can generate strong answers cheaply, the scarce skill becomes understanding, validation, and explanation thedataexchange.media/quick-take-2...
thedataexchange.media
When Everyone Is 10x More Productive, No One Is
Ben Lorica and Evangelos Simoudis on AI IP Protection, Navier-Stokes Controversy, Human-Reserved Roles, and Workplace Involution.
000
Gradient Flow @gradientflow.com · 12/09/2026
BBC documentary on 9/11 and Saudi Arabia 👀 youtu.be/JF8EFyGaGDM?...
youtu.be
9/11: The Saudi Connection - BBC World Service Documentaries
YouTube video by BBC World Service
000
Gradient Flow @gradientflow.com · 11/09/2026
Ethics.Dev 🎗️ The Daily Brief on AI and Society ethics.dev
ethics.dev
Ethics.dev
Tracking AI’s Impact on Society
000
Gradient Flow @gradientflow.com · 11/09/2026
Don’t use your phone in public in #London #UK youtu.be/DuzAe_jHcQw?...
youtu.be
Can London stop the phone snatchers?
YouTube video by BBC London
000
Gradient Flow @gradientflow.com · 10/09/2026
One stereo image may be worth 1,000 regular images for training AI world models. Stereo data gives you the geometry of a scene for free, no guessing required. Leia Inc.'s David Fattal explains why this changes everything for robotics and spatial AI thedataexchange.media/david-fattal...
thedataexchange.media
The Missing Data for Spatial Intelligence
David Fattal on Stereo Data, World Models, 3D Displays, and the Future of Spatial AI.
000
Gradient Flow @gradientflow.com · 10/09/2026
👉 gradientflow.substack.com/p/this-is-ho...
000
Gradient Flow @gradientflow.com · 09/09/2026
AI Self-Improvement Without the Hype 🎯 Most talk about self-improving AI jumps straight to recursive self-improvement. But there are really three separate ideas: continual learning, bounded self-improvement, and full RSI. They are not steps on a ladder 🎗️ gradientflow.substack.com/p/this-is-ho...
gradientflow.substack.com
Your AI model is a rental (but the loop is an asset)
Put data, machine learning, and AI to work.
000