
Building something new | Previously VP of Inbound at @ibm. Also like cooking, nba and other.
New York
Joined March 2010
- Tweets7.1K
- Following2.3K
- Followers3.2K
- Likes1.6K
I fear too many products right now are wrappers bridging claude & many 3rd party apis.
And the assumption is “no way these users 5 miles deep in ai psychosis will do all these integrations themselves.”
That’s not a good assumption. Claude maniacs have to do ~~something~~.
Trying to build something new but the gravitational force of building a slightly different harness is too strong, pulling us into a black hole “the winter batch is coming” you hear in your nightmares until reality collapses in on itself as you think… harness orchestrator?
I wonder how stable saas arr numbers will be in the ai era.
It is just so easy to migrate these days if you know your way around coding agents.
It used to be if you weren’t exhausted by your use of agents, you weren’t using them enough.
Now they are good enough to run autonomously for hours if you set it up right so if you are exhausted by your use of agents, you aren’t using them enough.
If you want to token max as a creator I suggest working on video.
Wow you can spend a lot of money really fast doing video.
Or I’m doing it wrong. Perhaps both.
If you are running a mostly autonomous fleet via orchestrator pattern, have your orchestrator agent just keep an “open decisions” artifact open that has explanations, this or that visual tradeoffs, etc.
So instead of 7 bullets you dont understand the context on or need explanations you just have a nice running artifact with a list of decisions.
I’m not optimistic around where the zero click version of search takes us.
It’s very different from ads or social where the work is seen by people but shows up as direct or branded search.
In a zero click search world humans aren’t reading the content.
The way you build and motivate a great team of writers and creators is with the fact that people read the work.
If no one reads the work, I don’t know how you get anyone to care about the quality of editorial. Which becomes this very destructive feedback loop.
Commercial landing pages are a different story. No one doing cro cares if people are reading their features blurb as long as you put conversions on the board.
I think there are two ways to approach this.
First is to simply give up. Which I’m increasingly of the mind is the right thing to do for many companies.
Second is to get better. The argument for this is hard but this is roughly how I think about…
- In general my pov on channels is not do they work but what does it take to get them to work. I dont like removing opportunities to reach the audience.
- If everyone else abandons these programs, it will become easier to win.
- Search editorial often wasn’t promoted across other channels for quality and timeliness reasons. “Making it better” gives search content the oppty to get multi-channel outcomes.
- People don’t think hard enough about how to amortize the cost of research over multiple pieces of execution. One deep dive on a topic can produce written content, youtube content, social content, newsletter content etc. It may actually be different content in each channel but the real cost of most work is not the writing, but the research, organizing your thoughts, etc. You can reduce the cost of search content by amortizing 80% of the expense across other pieces of execution.
- Top of funnel queries in ai engines definitely produce lower mention rates than any other part of funnel and the simplest expression of these questions often doesn’t trigger grounding. But that’s misleading. If it was true there’d be no ai referral traffic to evergreen informational content and thats just not the case. Probably because similar to trad search there’s 10x+ more informational queries (at least in b2b) vs commercial. Which means even if mention rates are 5x lower, its still a bigger channel by total mention volume (admittedly in a lower intent context).
- Perhaps most importantly, and in the context of “just make it better” I am honestly shocked that no one is putting all these claude max subs to good use and building incredible new experiences on the web, experiences that wrap this content in a totally new category of digital “product.” Experiences that can hunt in search but that attract attention from all channels.
- Do more research and produce more stats and data. Then your editorial becomes distribution for reports.
Even with all that it still might not work. But if I really pushed it and got nowhere, then I’d feel better at calling it dead dead, at least in the case of the given company.
One thing I will add is that there was a certain type of informational content people made that was purely about traffic.
Content that would have no value to a company outside the benefits of seo, which were already limited.
Google already hit many of these sites and content types hard.
That content now has negative value.
Costs money
Will not find an audience
Possibly damaging to the site
I’m not too upset about this particular behavior having reached the end of its shelf life.
One of the big changes for me is that the process of building software now resembles how I write.
Either ironic or obvious given the “hottest new programming language is English.”
But when I seriously write there’s a lot of words that get written and then rewritten, refactored or removed.
Because I’m finding my way to the central idea of a thing and getting there requires exploring many of its edges.
Product is like this too but because building used to be so expensive there was a forcing mechanism to explore the idea in documents or maybe design.
Now there isn’t.
I suspect now that a good way build software may be to build quite a bit more, explore the edges as working product and then refactor or delete things that aren’t working or that don’t feel organic to the product’s core.
However this part will require a different sort of organizational discipline and will become more challenging the larger the team is.
This type of work was never easy in large teams but things moved slowly enough where sometimes you could figure out how to align or refactor things before it was too late.
But there wasn’t an abundance of time and it took real effort to sometimes drive even small changes in alignment.
Now all these teams can move so fast that things are shipping faster than you can even talk about them.
It may be that culture becomes an even more important aspect of creating great product given how much harder it is organizationally to delete or refactor things than it is to build new things.
More realistically it may be the case that team size is inversely correlated with product quality and success.
It is hard enough to refactor your own thinking into something clean and coherent.
Doing it across a large org of slop cannons is not only intellectually and creatively challenging but also emotionally exhausting.
Someone behind the software
I think good products carry something of the people who make them. Their experience, what they care about, and how they think things should work. You might not know who those people are, but you can feel their perspective in the product.
The industry often talks about software as though it comes out of a factory. You collect requirements, build features, and measure whether they work. Follow the process and you should get a good result. Now we're talking automating this whole pipeline with agents.
But two teams can talk to similar customers, have access to the same technology, and make very different things. The difference comes from how they understand what they see and the style they go about building it. The founder background or company DNA also shapes this a lot.
At Linear, we have always valued intuition over process or data. I think people sometimes understand that as making decisions based on personal preference, without much thought or evidence. What we mean by it that your intuition develops through experience and understanding. You work on something, talk to customers, make mistakes, and learn what to pay attention to. The intuition is your point of view on something, and forcing to use your judgment, not hide behind faceless data.
Before we launched, we talked to people across companies and roles about how they worked. We had our own frustrations with this type software, but those conversations helped us understand what mattered beyond our own experience. Our perspective got better after those conversations.
So eventually the products develops a character through what its makers pay attention to, what they decide not to do and what they keep working on after it technically works. I sometimes think of this as the soul of a product. You feel that someone cares about how the whole thing comes together.
A feature can work perfectly well on its own and still make the product worse. Recognizing that requires some belief about what the product should be. Someone has to make the call, then stay close enough to customers to see whether they were right.
Recently, many products are starting to feel messy to me, including products from very successful companies. There seems to be more thought going into what the technology can do and what can be added than the form it should take. The product changes are primarily additive, not improving the core problem.
I think the reasons for this is that it's lot easier to add new capabilities, than rethink existing capabilities. So each product becomes a collection of abilities, without actually having unified point of view.
I wonder whether some of this comes from how the work is organized. Individuals work with agents to produce changes independently, each operating a small software factory. Each change might make sense on its own, but together they start to resemble Frankenstein’s monster rather than a carefully shaped sculpture.
Without the shared perspective,
soul, or thought, the product becomes an average of all those outputs. The choices that once gave it character are being diluted by day. More things work, but it becomes harder to feel someone or even something behind them.
Making software is getting easier. I hope that gives people more room to develop a perspective and follow it further. To understand a problem deeply and keep shaping an answer until the pieces belong together.
I want to feel the creator behind the software, not the factories.
Watching the broncos is a truly horrific experience.
It’s not an excuse to play like this but Bo is clearly physically not right. Those last two plays of OT were ones he makes last year. He can’t run anymore.
What are the use cases for where you want multiple humans in a chat with an agent or multiple agents and the venue is not slack or teams?
Personal agent factories are an intoxicating amount of leverage.
And to get the most out of it you have to exit the agent “toddler phase.”
My setup is now…
Orchestrator agent
Architect agent
Production design agent
Exploratory agent
10 build agents
The better agents have gotten the more I remove myself. They can now do everything without me besides deploying to prod.
Orchestrator has a 10 min ping routine where it checks if every other agent is busy/idle and if idle gets em going again.
All testing is now running remotely in a big linux box in the cloud.
We have “night mode” where computer is prevented from sleeping and no tasks that require my approval or feedback are worked by any lane (so no one gets blocked during that whole window).
Estimate this morning is that there’s 4 days of backlog fully scoped so theoretically I could just do nothing and these things would mostly just go for days while I did nothing.
The “toddler phase” was so exhausting because you could just never look away.
You’d look at tool calls and be like what are you doing.
The more agents you had running with restricted permissions the more it felt like the slot machine meme just hitting the same button 500 times a day.
You couldn’t do much “thinking” because you are just babysitting these things 24/7.
But if you get to a place where your core agents have a long backlog and are running largely autonomously?
Holy shit.
I can sit down and do lots of “real work” and not feel like I’m trading my strategy/thinking time for execution because a fleet of agents is just doing weeks of work every day without me.
The reason I got into “inbound” for GTM was because I was obsessed with the economics and leverage and compounding gains.
It could not be more clear to me that turning yourself into 10, 20, 100 yous is the next version of this.
What an incredible time.
Sessions were supposed to be cattle but they became pets.
Makes you wonder if session is even the right model.
Nevermind maybe a skill issue but Claude Code is like impossibly bad at this I don’t get it.
Big change to my core workflow today.
Claude design had a notification this morning it’s moving into artifacts.
I don’t know if artifacts always had these capabilities but I noticed a few days ago the main web app started doing a nice job building whole screens and flows.
And was doing them as part of nicer, more normal canvas like experience.
Maybe it could always do that and I only noticed because I started to build some workflows where my other claudes were briefing claude design more.
But then I realized you could operate artifacts from inside claude code on desktop app too.
Which means I can now have a design session run by claude code.
That session can communicate directly with all the other claudes in the fleet.
It can read/write directly to the main github repo.
It can walk the main app on local and take screenshots and stick them on a visual canvas.
Which means I can be like hey I wanna redesign this thing and claude can just put together a canvas with all the relevant current screens.
We can then iterate on everything in the canvas. And then write all the design changes directly to github and then communicate directly with fleet leader and architect to resolve any issues/questions.
So good. I love it.
Big change to my core workflow today.
Claude design had a notification this morning it’s moving into artifacts.
I don’t know if artifacts always had these capabilities but I noticed a few days ago the main web app started doing a nice job building whole screens and flows.
And was doing them as part of nicer, more normal canvas like experience.
Maybe it could always do that and I only noticed because I started to build some workflows where my other claudes were briefing claude design more.
But then I realized you could operate artifacts from inside claude code on desktop app too.
Which means I can now have a design session run by claude code.
That session can communicate directly with all the other claudes in the fleet.
It can read/write directly to the main github repo.
It can walk the main app on local and take screenshots and stick them on a visual canvas.
Which means I can be like hey I wanna redesign this thing and claude can just put together a canvas with all the relevant current screens.
We can then iterate on everything in the canvas. And then write all the design changes directly to github and then communicate directly with fleet leader and architect to resolve any issues/questions.
So good. I love it.
So many engineers feel like coding is over.
Meanwhile I feel like it is over for anyone that doesn’t become very comfortable with code.
Great vid on creator programs.
Not a lot that can make me stop scrolling and watch a 15 min video.





