@smithstephen.bsky.social
You found the 'training off' toggle. Good. But did you know the thumbs-up button can quietly override it? Here's the mental model I use to keep client work out of AI improvement pipelines.
www.smithstephen.com/p/you-turned...
www.smithstephen.com/p/you-turned...
You Turned Off Training. The Feedback Button Didn't Get the Memo.
Consumer AI privacy is a setting. Your feedback is an action. Guess which one wins.
www.smithstephen.com
December 24, 2025 at 1:59 PM
You found the 'training off' toggle. Good. But did you know the thumbs-up button can quietly override it? Here's the mental model I use to keep client work out of AI improvement pipelines.
www.smithstephen.com/p/you-turned...
www.smithstephen.com/p/you-turned...
There's a difference between looking busy and moving with direction. I've been watching OpenAI ship a lot lately, but Google and Anthropic feel like they know exactly where they're going. Here's why that distinction matters for your AI strategy.
www.smithstephen.com/p/openai-won...
www.smithstephen.com/p/openai-won...
OpenAI Won the Demo. Google and Anthropic Are Winning the Default.
When everyone's watching your magic tricks, it's easy to miss who's rewiring the building.
www.smithstephen.com
December 22, 2025 at 7:17 PM
There's a difference between looking busy and moving with direction. I've been watching OpenAI ship a lot lately, but Google and Anthropic feel like they know exactly where they're going. Here's why that distinction matters for your AI strategy.
www.smithstephen.com/p/openai-won...
www.smithstephen.com/p/openai-won...
"How do you keep up with all the AI news?" It's the question I get more than any other. Here's my honest answer: I don't keep up with all of it. I built a system that filters what matters for business decisions and ignores the rest. This is that system.
www.smithstephen.com/p/you-asked-...
www.smithstephen.com/p/you-asked-...
You Asked How I Keep Up With AI. Here's the Honest Answer.
Spoiler: I don't. Nobody can. Here's what I do instead.
www.smithstephen.com
December 21, 2025 at 2:56 PM
"How do you keep up with all the AI news?" It's the question I get more than any other. Here's my honest answer: I don't keep up with all of it. I built a system that filters what matters for business decisions and ignores the rest. This is that system.
www.smithstephen.com/p/you-asked-...
www.smithstephen.com/p/you-asked-...
Hanging out in the office couch with me while I work.
December 19, 2025 at 4:35 PM
Hanging out in the office couch with me while I work.
Google didn't win 2025 with one breakthrough model. They won by shipping AI into the places work already happens, at prices that make broad rollout possible. That changes everything for 2026 planning.
www.smithstephen.com/p/google-won...
www.smithstephen.com/p/google-won...
Google Won 2025 by Making AI Boring
Demos don't compound. Defaults do.
www.smithstephen.com
December 18, 2025 at 4:24 PM
Google didn't win 2025 with one breakthrough model. They won by shipping AI into the places work already happens, at prices that make broad rollout possible. That changes everything for 2026 planning.
www.smithstephen.com/p/google-won...
www.smithstephen.com/p/google-won...
Everyone has AI tools now. Almost no one has AI advantage. The difference isn't better models or bigger budgets. It's workflow redesign, scaling discipline, and measurement that actually matters. Here's what 2025 taught us about what 2026 will demand.
Everyone Has AI Now. Almost No One Has AI Advantage. What Actually Changes That.
Why your tools are fine, your workflows are the problem, and your board is about to notice
www.smithstephen.com
December 17, 2025 at 11:38 PM
Everyone has AI tools now. Almost no one has AI advantage. The difference isn't better models or bigger budgets. It's workflow redesign, scaling discipline, and measurement that actually matters. Here's what 2025 taught us about what 2026 will demand.
The AI gap isn't widening because some companies have better prompts. It's widening because some companies built a compounding system while everyone else kept running pilots. New piece on what the EY AI Pulse Survey actually means for your 2026 planning.
www.smithstephen.com/p/the-compou...
www.smithstephen.com/p/the-compou...
The Compounding Problem: How AI Leaders Are Pulling Away While Everyone Else Runs Pilots
EY's survey shows 96% of AI investors seeing gains. The uncomfortable part is what they're doing with them.
www.smithstephen.com
December 15, 2025 at 4:53 AM
The AI gap isn't widening because some companies have better prompts. It's widening because some companies built a compounding system while everyone else kept running pilots. New piece on what the EY AI Pulse Survey actually means for your 2026 planning.
www.smithstephen.com/p/the-compou...
www.smithstephen.com/p/the-compou...
OpenAI released ChatGPT 5.2 and I translated it into executive English. The short version: it finally remembers page one when it gets to page sixty, and that changes which workflows are worth automating.
www.smithstephen.com/p/chatgpt-52...
www.smithstephen.com/p/chatgpt-52...
ChatGPT 5.2: The Model That Finally Remembers Page One When It Gets to Page Sixty
Model numbers keep climbing. This time, the upgrade that matters is the one your contracts have been waiting for.
www.smithstephen.com
December 11, 2025 at 11:23 PM
OpenAI released ChatGPT 5.2 and I translated it into executive English. The short version: it finally remembers page one when it gets to page sixty, and that changes which workflows are worth automating.
www.smithstephen.com/p/chatgpt-52...
www.smithstephen.com/p/chatgpt-52...
Most executives assume their AI rollout is blocked by budget or technology. OpenAI's latest data suggests otherwise. The companies seeing 8x more usage aren't running better pilots. They're changing behavior, and the gap is widening faster than you think.
www.smithstephen.com/p/your-ai-li...
www.smithstephen.com/p/your-ai-li...
Your AI Licenses Aren't the Problem. Your Culture Is.
A million businesses have access to the same AI. Most of them are still stuck at the pilot stage.
www.smithstephen.com
December 8, 2025 at 10:34 PM
Most executives assume their AI rollout is blocked by budget or technology. OpenAI's latest data suggests otherwise. The companies seeing 8x more usage aren't running better pilots. They're changing behavior, and the gap is widening faster than you think.
www.smithstephen.com/p/your-ai-li...
www.smithstephen.com/p/your-ai-li...
I tested Gemini with a messy PDF, two dashboard screenshots, and a question that would normally go to an analyst. It answered like a smart colleague who'd actually read the footnotes. That's the multimodal shift: AI that sees your work as it actually exists, not as a cleaned-up summary.
The End of Copy-Paste AI: What Happens When Models Actually Look at Your Work
Gemini 3 Pro just looked at your 90-page contract and had thoughts.
www.smithstephen.com
December 7, 2025 at 10:33 PM
I tested Gemini with a messy PDF, two dashboard screenshots, and a question that would normally go to an analyst. It answered like a smart colleague who'd actually read the footnotes. That's the multimodal shift: AI that sees your work as it actually exists, not as a cleaned-up summary.
Every month this year, something launched that would have been front-page news in 2024. I wrote the executive summary: which releases changed the game, which ones are noise, and the four things to do before Q1 planning kicks off.
www.smithstephen.com/p/2025-the-y...
www.smithstephen.com/p/2025-the-y...
2025: The Year AI Stopped Demoing and Started Delivering
Because your team is still explaining AI with 2023 examples, and that's starting to show
www.smithstephen.com
December 5, 2025 at 5:02 PM
Every month this year, something launched that would have been front-page news in 2024. I wrote the executive summary: which releases changed the game, which ones are noise, and the four things to do before Q1 planning kicks off.
www.smithstephen.com/p/2025-the-y...
www.smithstephen.com/p/2025-the-y...
I built a fully functional software agent in ten minutes without writing a single line of code. The real question isn't whether this is possible; it's whether your organization is ready for what happens when every employee can do the same.
www.smithstephen.com/p/the-it-tic...
www.smithstephen.com/p/the-it-tic...
The IT Ticket Is Dead: How AI Just Made Every Employee a Developer
No Jira ticket. No sprint planning. No six-week estimate. Just a working solution.
www.smithstephen.com
December 2, 2025 at 4:36 PM
I built a fully functional software agent in ten minutes without writing a single line of code. The real question isn't whether this is possible; it's whether your organization is ready for what happens when every employee can do the same.
www.smithstephen.com/p/the-it-tic...
www.smithstephen.com/p/the-it-tic...
Ilya Sutskever helped build GPT. Now he's running a $5 billion startup betting that "more GPUs" won't get us to AGI. His thesis explains why your AI tools crush benchmarks but stumble in real workflows, and what leaders should do about it.
www.smithstephen.com/p/the-man-wh...
www.smithstephen.com/p/the-man-wh...
The Man Who Built GPT Says More GPUs Won't Get Us to AGI
His new $5 billion company is betting on something else entirely
www.smithstephen.com
December 1, 2025 at 10:08 PM
Ilya Sutskever helped build GPT. Now he's running a $5 billion startup betting that "more GPUs" won't get us to AGI. His thesis explains why your AI tools crush benchmarks but stumble in real workflows, and what leaders should do about it.
www.smithstephen.com/p/the-man-wh...
www.smithstephen.com/p/the-man-wh...
Everyone's talking about McKinsey's 57% automation number. Almost no one is reading what the report actually says. Here's the operator's guide to what matters and what's noise.
www.smithstephen.com/p/57-of-work...
www.smithstephen.com/p/57-of-work...
57% of Work Hours Could Be Automated. So Why Aren't They?
A guide to reading the new McKinsey Global Institute report like an operator, not a headline writer.
www.smithstephen.com
November 30, 2025 at 9:37 PM
Everyone's talking about McKinsey's 57% automation number. Almost no one is reading what the report actually says. Here's the operator's guide to what matters and what's noise.
www.smithstephen.com/p/57-of-work...
www.smithstephen.com/p/57-of-work...
I gave Claude Opus 4.5 a 50-page PDF and asked for a Board deck. Two minutes later, I had a downloadable PowerPoint with ten slides, working charts, and structure that made sense. This changes everything about how AI fits into actual work if you live in a world that revolves around Microsoft Office.
I Gave The New Claude Opus 4.5 a 50-Page PDF. It Handed Me Back a Board Deck.
Anthropic finally figured out what we actually do all day
www.smithstephen.com
November 25, 2025 at 3:28 PM
I gave Claude Opus 4.5 a 50-page PDF and asked for a Board deck. Two minutes later, I had a downloadable PowerPoint with ten slides, working charts, and structure that made sense. This changes everything about how AI fits into actual work if you live in a world that revolves around Microsoft Office.
Nano Banana Pro is a ridiculous name for what might be one of the most important AI releases of the year. Google buried a thinking model inside an image tool, and most businesses won't realize the implications until their competitors do. Full breakdown.
www.smithstephen.com/p/google-jus...
www.smithstephen.com/p/google-jus...
Google Just Showed Us What AGI Will Look Like. It's Disguised as an Image Generator.
Nano Banana Pro doesn't guess. It reasons, critiques, and corrects itself before rendering a single pixel.
www.smithstephen.com
November 24, 2025 at 5:29 AM
Nano Banana Pro is a ridiculous name for what might be one of the most important AI releases of the year. Google buried a thinking model inside an image tool, and most businesses won't realize the implications until their competitors do. Full breakdown.
www.smithstephen.com/p/google-jus...
www.smithstephen.com/p/google-jus...
Your team is probably asking about Gemini 3 right now. The benchmarks look decisive, the demos are impressive, and Google's betting big on agents. But the real question isn't whether Gemini 3 is powerful. It's which model stops being your backup and becomes your default.
Gemini 3 Is Probably The Best Model Right Now. That's Not The Question.
The real decision isn't about power. It's about where hallucinations will cost you money.
www.smithstephen.com
November 20, 2025 at 5:16 AM
Your team is probably asking about Gemini 3 right now. The benchmarks look decisive, the demos are impressive, and Google's betting big on agents. But the real question isn't whether Gemini 3 is powerful. It's which model stops being your backup and becomes your default.
Most AI tools are really good at giving you more work disguised as help. Google’s new Gemini 3 Pro is betting on something different: AI that actually finishes tasks instead of just starting them. I wrote about what that looks like for your team and your operations.
Google Finally Built an AI That Finishes the Job
Your team has been drowning in AI suggestions. Google's betting you'd rather have AI execution.
www.smithstephen.com
November 19, 2025 at 12:18 AM
Most AI tools are really good at giving you more work disguised as help. Google’s new Gemini 3 Pro is betting on something different: AI that actually finishes tasks instead of just starting them. I wrote about what that looks like for your team and your operations.
Most people treat AI like a search bar. The ones getting real leverage treat it like a workshop full of tools where they're the architect. After thousands of hours of real-world use, here's the difference between one-off answers and systems that compound.
www.smithstephen.com/p/ive-used-a...
www.smithstephen.com/p/ive-used-a...
I've Used AI Every Day for Three Years. Here's What Nobody Tells You About Making It Work.
The secret to AI leverage: treating it like a slightly overeager junior analyst, not a crystal ball.
www.smithstephen.com
November 18, 2025 at 7:42 AM
Most people treat AI like a search bar. The ones getting real leverage treat it like a workshop full of tools where they're the architect. After thousands of hours of real-world use, here's the difference between one-off answers and systems that compound.
www.smithstephen.com/p/ive-used-a...
www.smithstephen.com/p/ive-used-a...
Your team is comparing Google's Gemini against OpenAI and Anthropic. But you're probably asking the wrong question. The real story isn't which model is 2% better this month - it's why Google suddenly has the organizational structure to keep improving while everyone else is scrambling.
Google's Secret Weapon Isn't Gemini - It's the Stubborn London Lab That Built It
What happens when you let researchers ignore quarterly earnings for a decade (spoiler: Nobel Prizes and actual competitive advantage)
www.smithstephen.com
November 16, 2025 at 5:42 AM
Your team is comparing Google's Gemini against OpenAI and Anthropic. But you're probably asking the wrong question. The real story isn't which model is 2% better this month - it's why Google suddenly has the organizational structure to keep improving while everyone else is scrambling.
Dashboard debates eat weeks and produce nothing. I tested three LLMs on the same Series A brief - one delivered a board-ready mock, two didn't - plus the exact prompt to try Monday.
www.smithstephen.com/p/stop-debat...
www.smithstephen.com/p/stop-debat...
Stop Debating Your Dashboard - Prototype It in Two Hours
I fed the same prompt to three LLMs and got three different executive dashboards. One was board-ready, one was getting there, one missed the mark. Here's the breakdown and the exact prompt.
www.smithstephen.com
November 11, 2025 at 11:32 PM
Dashboard debates eat weeks and produce nothing. I tested three LLMs on the same Series A brief - one delivered a board-ready mock, two didn't - plus the exact prompt to try Monday.
www.smithstephen.com/p/stop-debat...
www.smithstephen.com/p/stop-debat...
Dirty dog then clean dog. Magnus LOVES getting a bath.
November 6, 2025 at 7:14 PM
Dirty dog then clean dog. Magnus LOVES getting a bath.
While you're still explaining to your team what ChatGPT is, someone in your industry is running automated competitive analysis every morning before their first coffee, building financial models through conversation, and creating presentation-ready visualizations without touching PowerPoint.
The Top AI Capabilities Every Executive Should Be Using Right Now (But Probably Aren't)
Why learning to build apps in ChatGPT, automate spreadsheets in Claude, and schedule research in Gemini matters more than understanding how LLMs work.
www.smithstephen.com
November 5, 2025 at 2:39 AM
While you're still explaining to your team what ChatGPT is, someone in your industry is running automated competitive analysis every morning before their first coffee, building financial models through conversation, and creating presentation-ready visualizations without touching PowerPoint.
Some teams are still scheduling "deck week." Others type a prompt in Gemini, export to Slides, and spend that week sharpening strategy instead. The time asymmetry is creating a capability asymmetry.
www.smithstephen.com/p/from-gemin...
www.smithstephen.com/p/from-gemin...
From Gemini Prompt to Google Slides in 10 Minutes. Your Team Still Needs 10 Hours. Here's Why That Matters.
When competitors draft three narrative approaches in the time your team builds one slide deck, that's not a productivity gap, it's a strategic capability gap.
www.smithstephen.com
November 4, 2025 at 5:55 AM
Some teams are still scheduling "deck week." Others type a prompt in Gemini, export to Slides, and spend that week sharpening strategy instead. The time asymmetry is creating a capability asymmetry.
www.smithstephen.com/p/from-gemin...
www.smithstephen.com/p/from-gemin...
If you are in the LA Area on November 5, please join Lauren Eve Cantor and I for AI in Action....sign up link in the comments
November 3, 2025 at 12:33 AM
If you are in the LA Area on November 5, please join Lauren Eve Cantor and I for AI in Action....sign up link in the comments