Shawn Scully

AI product guy | ex-AIML @Apple, Turi (acq.  ) | founder | physics @Stanford, @Cornell | from Maine in Seattle | I post on AI, product, yurts, & tasty things.

[Shawn Scully @maketaster]

The AI discourse has mode collapsed.

The same loops on repeat: AI takes all jobs. Who's winning the AGI race. AI will cure disease… and aging. New model drops, gossip cycle, repeat.

100M brains interconnected, reinforced by each other's outputs, losing signal with every pass. The tails of the distribution are getting lost.

Stuck in a giant game of telephone. Confusing it for insight. Social model collapse.

[Shawn Scully @maketaster]

inspiration.

[Shawn Scully @maketaster]

This

[OpenAI @OpenAI]

GPT-6 and Intelligent UI, now rolling out in ChatGPT for everyone.

Intelligent UI in ChatGPT delivers fast, interactive answers that make everyday questions more visual, complex topics easier to grasp, and interactive tools for your task available on the spot.

[Shawn Scully @maketaster]

right

[Shawn Scully @maketaster]

the

[Shawn Scully @maketaster]

is

[Shawn Scully @maketaster]

Unleash the builders.

[Robert Scoble @Scobleizer]

My feed is of all sorts of people building all sorts of wild contraptions.

I love my feed today.

[Shawn Scully @maketaster]

@AdamWoptimized Inter-active. Damn!

[Adam Wolnikowski @AdamWoptimized]

Decision models can play Dark Souls in real time!

I’ve been trying to better understand the spatial/temporal reasoning and latency tradeoffs of different models, so I had them all fight Gundyr.

GPT-6 Luna Decisions came out on top, with multiple no hit runs:

[aviel @aviel]

At this point I’ve heard people declare that B2B SaaS is dead more times than I can count.

And there is some truth to it.

When I built Root, the AI agent that manages the Foundations community, we built in-house instead of buying SaaS (we did of course buy some platform level SaaS for hosting, email infra, etc). We created exactly what we needed rather than assembling a collection of products that were individually configurable but collectively disconnected.

It worked extremely well. Our member experience is also top notch vs. the alternative.

But building software from scratch is still a lot to ask of most companies. Especially if you want it to power a large part (or all) of your business. At the same time, the alternative is often worse: buying a configurable SaaS product that never quite fits, or stitching together several products that never work together as one system.

So what comes next?

My current answer is thin extensible platforms that live at the right layer of abstraction for a category of company.

They should provide enough structure that you are not starting from zero (or near zero like a horizontal platform), while remaining flexible enough to become the software that your business actually needs.

But this is not just another VC thought piece… I tested the theory by doing the work.

I extracted the reusable layer from Root to an agent platform for communities, separated it from the Foundations-specific application, and rebuilt our company specific data structures and workflows on top of it.

I do not know whether I will ever release the platform publicly. But it is real, complete, and working. More importantly, it bolstered my confidence that whatever comes “after” SaaS will live at this layer, not below and not above.

You can check it out at https://t.co/yXnfQypENV

But there is a problem…

A thin, extensible platform isn’t that valuable on its own. You still have to define your own workflows, rules, context, and desired outcomes. Increasingly, though, much of that work is just writing, adapting, and composing prompts rather than traditional software development. But because the platform is not solving the entire end-to-end problem, it can’t command the same pricing as traditional SaaS. In other words, it’s highly likely that this is indeed a real category, just not one that is venture fundable.

[Weave Robotics @weaverobotics]

Today, we’re launching our home robot Isaac 1.

Isaac 1 deliveries will begin this fall.

Order yours below.

[Shawn Scully @maketaster]

@hamburger It must mean they are switching to a b2b co.

[Shawn Scully @maketaster]

@davidhoang Renaming the product could be one of the best enterprise product marketing / sales enablement moves they could do.

[Shawn Scully @maketaster]

@signulll + -AI-startup-bigP-product-guy-with-yurt perspectives.

[Shawn Scully @maketaster]

@signulll Because I need friends that are not bots.

[Shawn Scully @maketaster]

@ChaiKamNamak @AdrienLE Fair point!

[Shawn Scully @maketaster]

@grok @AdrienLE There ya go!

[Shawn Scully @maketaster]

@AdrienLE @grok Sorry, point @grok, what is “yeeting”?

[Shawn Scully @maketaster]

@AdrienLE Haha, well yeeting @grok ?

[Shawn Scully @maketaster]

@rcmisk So I guess the idea is to get 1 person you think is target to give feedback. Then iterate. 5 get feedback. Iterate. 10. 100. 1000.