Todd Geist

It’s terrible,

But maybe it is necessary. Maybe new systems can rise out of the compost pile of all the old systems that have been enshittified.

SaaS is dead. It will take 10 yrs to die, but it’s done

Looks like an American Kestral maybe.??

Super cool

Excellent, and thought provoking.

I dunno, it gives me the "I need a whole internet to believe I am amazing, just to drag my ass out of bed in morning vibe"

"Pathetic" is how I would put it

meanwhile IRL

www.usatoday.com/story/money/...

Amazon blocking Perplexity isn't about transaction rules. It's about survival.

Shopify's bet on agent commerce isn't just strategy—it's an existential threat to platform monopolies everywhere.

This threatens every platform built on attention mining—Amazon, Google, DoorDash. The internet consolidated among 5-10 companies. Agent networks could reverse that through infrastructure players with asymmetric incentives.

Shopify says AI traffic is up 7x since January, AI-driven orders are up 11x | TechCrunch Shopify is bullish on AI-powered shopping agents, citing AI as an "incredible tool" to enable more entrepreneurs and calling it the "biggest shift in technology since the internet" during its third-qu...

The proof: Shopify open-sourced the Agentic Commerce Protocol with OpenAI and Stripe. Partnered with Perplexity (yes, the company Amazon blocked). Results: 7x AI traffic, 11x orders since January.

techcrunch.com/2025/11/04/s...

Unlike Amazon, Shopify is infrastructure, not a destination. They profit when merchants succeed—regardless of which interface drives the sale.

While Amazon blocks agents, Shopify is building the rails for them.

Platform monopolies run on attention mining: capture users in a destination → steer them toward max platform profit → consolidation.

AI agents eliminate the destination entirely. No destination = no attention to mine. The business model collapses.

AI agents want power. Platforms won't give it up Amazon vs Perplexity is the opening shot in a larger war

Amazon blocked Perplexity's shopping agent. @dfriedman nailed why: "Enclosure and abstraction cannot peacefully coexist."

But there's more to this story—and Shopify is making a massive bet on it.

davefriedman.substack.com/p/ai-agents-...

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

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.

He seems to be drawn to tacky gold stuff. Maybe this is the staff trying to help him find his way back, when he gets lost.🤪

Well they haven’t cut them out “entirely”. I am still going to use switch to sonnet or gpt 5 for some tasks, but the majority of the tokens I am burning are now with composer

Cursor spent 2 years using Anthropic and OpenAI models, collected massive amounts of data and use cases, then trained their own frontier-level model and cut both out entirely.

How is this not the playbook every AI application builder will follow?

Composer: Building a fast frontier model with RL · Cursor Built to make you extraordinarily productive, Cursor is the best way to code with AI.

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

I haven’t thought of that movie in 30 plus years, but I knew it instantly from that one shot. And then I remembered the skeleton fighting scene.

It is really fast, and so far the quality seems pretty good.

And being fast is pretty helpful. It’s easier to stay focused on what’s happening, which makes it easier to catch when it makes mistakes

Another very important fact that falls out of this is

You still have to use your brain.

It also explains why some people experience bigger wins than others. Some people’s bandwidth was much higher than others. They could type faster and the needed less googling to make shit work.

Those folks don’t get as much out of it as someone like me, who still can’t type with more than 2 fingers

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.

Ok I think “the Studio” is not good

ProofChat - Full featured AI Powered Chat embedded in and connected to your FileMaker applications. - www.proofchat.ai
www.youtube.com/watch?v=uZP8...

Thanks for mentioning this. I experience a lot of success.

And if you out there reading this do as well. It's ok. You don't have to feel bad. Other people get to have their own experiences. Just as you do.

Choose joy. And let others do the same.

For me, today, AI Coding brings me a lot of joy!

great article on working with Cursor or other AI coding tools.

anyblockers.com/posts/workin...

But if your chat app, like Claude or Goose, uses tools from an MCP Server, and it fails 1 in 4 times, you can correct it. You can tweak the context and try again.

YOU are the planning and correcting agent.

It's not a perfect system, but it can result in workflows that feel fast and effective.

this becomes even more obvious when you try to build multi-agent systems. If you have three agents that each fail 1 out of 4 times, your system is will only work 4 in 10 times, as the errors compound. .75 x .75 x .75 = .42

If each of these agents hides it memory the system as a whole can't learn.

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.

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.

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

😜