Todd Geist
banner
toddgeist.me
Todd Geist
@toddgeist.me
Among other things, building @ottomatic.cloud with TypeScript, Supabase, FileMaker, and an awesome team at proofgeist.com

Board member onesparkacademy.org and pacificzen.org.

Transformation happens in the dark!
The US AI strategy appears to be light a few trillion dollars on fire.

China's strategy appears to be do way more with way less.

x.com/Kimi_Moonsho...

I'd like to believe that American innovation goes beyond financial speculation and vibes. but it's getting harder to believe.
Kimi.ai on X: "🚀 Hello, Kimi K2 Thinking! The Open-Source Thinking Agent Model is here. 🔹 SOTA on HLE (44.9%) and BrowseComp (60.2%) 🔹 Executes up to 200 – 300 sequential tool calls without human interference 🔹 Excels in reasoning, agentic search, and coding 🔹 256K context window Built https://t.co/lZCNBIgbV2" / X
🚀 Hello, Kimi K2 Thinking! The Open-Source Thinking Agent Model is here. 🔹 SOTA on HLE (44.9%) and BrowseComp (60.2%) 🔹 Executes up to 200 – 300 sequential tool calls without human interference 🔹 Excels in reasoning, agentic search, and coding 🔹 256K context window Built https://t.co/lZCNBIgbV2
x.com
November 7, 2025 at 8:00 PM
I’ve used Cursor’s Composer—it’s legitimately good. Easily on par with Sonnet 4/4.5 and very fast. Makes me wonder if there’s any moat left for frontier model builders, and might explain why they’re trying to set $1.4T on fire.

cursor.com/blog/composer
Composer: Building a fast frontier model with RL · Cursor
Built to make you extraordinarily productive, Cursor is the best way to code with AI.
cursor.com
November 2, 2025 at 11:53 PM
I’ve said it before and I’ll say it again.

The real unlock in this generations AI models is higher Bandwidth

AI radically increases the bandwidth between your brain and the computer. All the wins stem from that.

All the “the shit doesn’t work” takes stem from missing this point.
October 15, 2025 at 4:16 PM
Ok I think “the Studio” is not good
October 10, 2025 at 2:42 AM
ProofChat - Full featured AI Powered Chat embedded in and connected to your FileMaker applications. - www.proofchat.ai
www.youtube.com/watch?v=uZP8...
ProofChat - AI Chat for FileMaker
Seamlessly integrate AI chat into your FileMaker applications. Query data, generate insights, update records, and create powerful workflows with natural language.
www.proofchat.ai
October 10, 2025 at 2:15 AM
great article on working with Cursor or other AI coding tools.

anyblockers.com/posts/workin...
Working with coding agents / anyblockers
What I have observed working well with AI coding agents during 2025 and practical tips to get the most out of them
anyblockers.com
October 9, 2025 at 6:15 PM
This article by @goldengatesteve is right on: fully autonomous agents aren't ready for prime time yet.

secondthoughts.ai/p/gpt-5-the-...

BUT Claude or Goose plus MCP servers can be very useful. Part of the reason is that you, THE HUMAN, are kept in the loop. That helps the convo stay on track.
GPT-5: The Case of the Missing Agent
AI has made enormous progress in the last 16 months. Agentic AI seems farther off than ever.
secondthoughts.ai
September 4, 2025 at 11:48 PM
writing code is about as "digital content creation-y" as it gets. so I guess I won't have to pay taxes on my wage... er "tips" I get every 2 weeks

😜
BREAKING: "Digital content creators" are among the dozens of occupations that the Treasury Department has deemed eligible for the "no tax on tips" deduction, per BI.

They explicitly name streamers, influencers, podcasters
September 3, 2025 at 10:56 PM
chii is pretty cool. Remote debugging with Chrome Dev tools
www.npmjs.com/package/chii

Works great for embedded web apps, like FileMaker web viewer apps like ProofChat and OttoDeploy. Pretty nifty

www.youtube.com/watch?v=gTKs...
Remote Debugging in Web Viewer
YouTube video by ProofGeist
www.youtube.com
September 3, 2025 at 10:47 PM
Misconception alert! 🚨 When you send images to OpenAI models like GPT-4, they're not just doing OCR—they actually understand the image. Also just cramming the base64 encoded image in the user prompt won't work. Here is how you do it with FileMaker's Generate Response From Model
youtu.be/1WUoYg6QdEg
Sending Images to OpenAI With Generate Response From Model
YouTube video by ProofGeist
youtu.be
August 14, 2025 at 9:23 PM
Day 1 of my FileMaker AI Livestream

www.youtube.com/watch?v=t9dr...
FileMaker 22 and AI on Pro Todd Geist - Day 1
YouTube video by FileMaker Training Videos
www.youtube.com
August 7, 2025 at 11:17 PM
Then I joined Ronnie and the RCC FileMaker Tv crew for a recap and Q and later that day.

www.youtube.com/watch?v=5Vwk...
July 14, 2025 at 3:06 PM
Claris FileMaker 2025 is out with some pretty cool new AI features. I joined the Claris team to talk about some of the new features and show some of the stuff we have been working on at Proof+Geist

community.claris.com/en/s/externa...
Claris Community (English)
community.claris.com
July 14, 2025 at 3:04 PM
How long before LLMs become LTMs?

Large Thought Models

It’s all about the IO. How fact can I get thoughts into the model and how fast can I get them out
July 8, 2025 at 7:25 PM
The strange little side effect of using LLMs is that I no longer even wanna type. I just wanna speak to the applications that I’m using and let the LLMs fix all of the punctuation errors and grammar errors.

It’s just way faster. High IO for the win!
July 8, 2025 at 3:00 PM
“I think I am the sand, man”

Ritchie
The Bear
July 7, 2025 at 2:06 PM
Here is another way to think about what AI is doing to the software industry.

Salesforce and all other SaaS companies have been charging rent on work that they did in the past. You paid a fee to get to use the particular combination of bits that make up salesforces source code.

Those days are over
June 4, 2025 at 7:30 PM
This is really good. It’s a way better, funnier write up of the way I have been feeling.

AI assisted coding is extremely good right now, & getting better all the time. If you are a software dev who doesn’t think so, I am sorry I don’t believe you have actually tried it.

fly.io/blog/youre-a...
My AI Skeptic Friends Are All Nuts
My smartest friends have bananas arguments about LLM coding.
fly.io
June 3, 2025 at 1:48 AM
It is now faster and easier to write a node.js script to remove the white background from an image from scratch than it is to remember how to use an app with a GUI that has that as a feature.
June 2, 2025 at 3:57 PM
@leerob.com

Running an MCP server on Vercel using the Vercel Mcp adapter. Connecting via Claude over http.

It all works great in Claude. 😍

But I am getting killed on compute.😟

How do I get the connection to close?

Each request stays open for the full 800 seconds.

Why?
May 29, 2025 at 7:11 PM
Write a note, let the studios know they made a mistake. Let the cast and crew know how much your loved the show.

savewot.com
Save The Wheel of Time - Finish the Story
SaveWOT is a fan-run website dedicated to supporting the effort to get The Wheel of Time television series back on a streaming service for a fourth season and beyond.
savewot.com
May 25, 2025 at 8:38 PM
I feel this. MCP is a protocol

With a collection of MCP servers connected to an MCP client, like Claude ( for now ), suddenly my stuff is coming to me. I don’t have to reach out to other apps. I can easily compose services and data from different apps on the fly. 🪄

www.anildash.com//2025/05/20/...
MCP is the coming of Web 2.0 2.0 - Anil Dash
A blog about making culture. Since 1999.
www.anildash.com
May 25, 2025 at 4:10 AM
Another hot take on where AI is taking the coding industry.

Licenses and IP based on code will increasingly be worthless.

AI can already code just about anything

Working code is not a moat anymore.

Proprietary data and RUNNING code that can do difficult things cheaply are the new moats
May 25, 2025 at 2:22 AM
All of the major AI model makers released new coding tools and or better coding models last week.

Now these tools can run for long periods of time, hours even, in the background.

It’s not going to be like paired programming anymore.

It’s going to be more like managing a team.
May 24, 2025 at 1:42 PM
The main problem with content is that it only needs to be harvested once and then it can be resold over and over again.

But apis can charge per use, because they are doing something other than just pointing at a resource.

I wonder if that becomes a thing. A new kind of Content behind an API?
There is just no way that the trade traffic for content economic engine that drove the last 20 years of the internet revolution survives.

AI search has already killed it. It just hasn’t figured out it is dead.

What comes next?

Ads in Chat will likely be tried. Not sure it will work.
May 18, 2025 at 5:13 PM