What’s Cookin’ in Hook-land? (Vol.9)

Hey There! Below is our/my weekly newsletter on anything integration and APIs. As always, please reply and let’s talk about whatever catches your attention, or any questions you might have.
​
Trying something new this week. New-ish. No images or links. Let me know if that's better or worse for you?

On a personal note...

Okay, let's just dive into a short story. Ollie rode his bike today for the first time. I was sitting at the office doing my thing; I think getting off of back-to-back calls. I get a video image. And a single line text of "OMG".

The context is that back in February, we had bought Ollie a bike we knew was too big for him. He's consistently been in the 95% or above percentile for height and weight, but also consistently been catching up to his body type. What I'm saying is: we bought big because he's big, but we knew he'd take some time to warm up into it. That plus he needed another few inches (and no snow on the ground) to get there.

Fast forward a few months, he'd gotten big enough to at least use it as a glider bike. Toes pushing along as he glided in front of us, now at the point of stopping at driveways and being alert... aka we're comfortable with him just cruising.

I tried once a few weeks ago to put the pedals on. And boy, let me tell you... 5 minutes in and I was exhausted. Also my lower back hurt from bending. Either parents way underplay the pain of helping their kids, or Ollie is just in an awkward height, or I now understand that training wheels are back-saving devices well worth their $30 price tag 😂

Anyway, no dice. He was not groking that he could just still glide and balance the way he was doing without the pedals. They threw off his game. So we removed them. And for the past few weeks, he's been going faster and faster, picking his feet up and putting them on the pedal stubs (to just rest, but... no idea if he was putting things together).

I guess this is a long winded way of telling you that this kiddo managed to figure out how to ride within 5 minutes of trying again. Nicole came outside and he was holding the pedals, asking her if she could put them on and ride.

She's now 20 weeks into pregnancy so she's staring at the pedals, calculating in her head what's being asked, and tentatively agrees expecting to be exhausted after 3 minutes of chasing/stabilizing.

None of that needed.

He got on. Was reminded that the balance was exactly the same as without pedals.

And he was off. No falls yet either!

What's this got to do with anything?

Nothing. Just memorializing a great moment for me and the family.

Also, I should note this is par for the course for him. He's regularly sort of just... watched. Practiced in different ways, mulled over it. And then when he tries something he picks it up quick, as long as he's given that runway to get his head around it first. Potty training was the same way. Writing. I think reading will go that way too.

Scout, on the other hand... that dude's just gonna learn by falling so many times 😂

Anywho, is that like you with your AI experiments? Are you Ollie, watching everyone do their thing, tentatively trying but not going to dive too deep, and then voila you get great results?

Or are you more like Scout, just going to try hard and keep failing until it finally clicks?

More on that later. Hope that inspired you to hop on a bike!

Best,

~Sean

What we're seeing

More LLM vulnerabilities (really, it's infinite)

I remain spooked about the vulnerabilities for LLMs, and haven't seen anyone provide a compelling safe case yet or system. I'll keep using personally, we'll use it for local development tasks, but for now we'll keep AI either tightly scoped in flows or close to the chest with human in the loop.

The MIT study for AI

You've heard about it. You've heard the hard takes in either direction. I stan with the middle crowd, who saw the note that personal adoption was actually helpful for productivity, while enterprise-wide initiatives have a high failure rate. Not surprising there, because that sounds like R&D and trying to put something in that has a high variance for being correct, with a low threshold of acceptance for enterprises. Just, no surprise.

If we were advising, we'd say "expect failure, learn from it, and benchmark what you're looking for so that when news comes out about the next set of models or systems, you'll know exactly when to do your thing". That's basically how we were treating AI at Left Hook for 3+ years, even prior to the LLM boom. I remember it was the moment structured outputs hit that I thought "okay, now it's possible, everything we've been thinking". Then when code editors went from having you copy and paste to actually suggested inline edits and letting you accept/reject, that was another step change moment.

My advice is to explore, figure out your step change needs and tests, then save those tests and every few months or releases, run your tests.

And all the while, reach for it for your personal needs. And don't get too sucked in, type your stuff. (Like this email. 95% typed by me)

Interview with Bret Taylor and Clay Bavor on Acquired

Former co-CEO Slack, etc. etc.) and Clay Bavor (20 years product at Google) with the Acquired podcast guys (ACQ2).

I dunno, I thought I'd have a bias against Bret because someone at that level has to be a bad person somehow. But he sounds great, he sounds like a nerd, and his cofounder sounds the same. And, they are being super pragmatic, which is inspiring. They are not trying to tackle the ocean. They're happy fishing in the pond until they grow to be the biggest fish before hopping out. These two wildly successful folks!

"Niching down", basically. I really appreciated their take on leveraging AI, and how they thought about what part of the stack to own. Let's make that a normal thing for people.

What we're hearing

Integrations are still the moat? And still hard to build?

We're kicking off a cohort of the unlimited integrations offer, and it's been interesting to hear the same pain points from different angles. And validating! The model so far has resonated because we're not promising X dollars for a given integration. We're promising a team of dedicated experts that are going to sprint at a healthy number of integrations, then like a steady drumbeat work to iterate with you over time.

That's just not what you're buying from anywhere else in the market. And when you try to stick your team on that internally, it's rare you can stick a team JUST on integrations for their monthly responsibilities. And then, you don't have a team that knows all of the different integrations that we know.

Sorry, not trying to advertise. Just trying to hit on the themes we're hearing when we talk to people, and this one is regular. I really appreciate when folks know how to use that expertise, and we're really happy to help them out.

While I may feel like Neo (I see everything as an integration like Neo seeing everything as code)... most people see apps and jobs to be done

Conversations about what we do, what I do, and how we do it and why it matters have come up frequently (not specifically with clients, thankfully 😂). I'm starting to get better at explaining it. But also what I'm realizing is that I'm probably the only person in my immediate circle who's tickled by the idea that "everything is integrations, all products are integrations". Everyone else? They just want to know what an app does for them, and then otherwise they want to move on with their days :-)

I think they FEEL the impact of well built and used integrations on a daily basis, but you don't SEE it happening. Which is by design.

Anyway, I'm probably not going to do a great job at convincing you to geek out about it. But, I'd just suggest you take a look around your life and do a 30 minute review of what things you use and do are integrated and automated. Especially across "app or system boundaries". Does your watch sync to your phone with health stats? Are you getting a Slack alert daily with your to-do list? Do you get an email notification with a weekly report about something? How about billing notifications? I guarantee you nobody these days is hand-pulling your report and sending you the email from your bank saying the monthly statement is ready.

It's everywhere. It's helpful to occasionally just, know about it.

What we're doing

Mentioned above, kicking off the first cohort

Even just going through the proposal and agreements was helpful, and we're planning on learning a lot through the next few months. The first thing that's standing out is that we have a process of understanding a product and a company. We have a set of questions. And it's always, always helpful to have a product owner walk us through what they think about the product, how to use it, and where they see integration value.

So that's one- we need a few days to really dig into a product and understand it.

Another is getting into the tech stack. That often helps inform past decisions, current UX/UI behavior and constraints, and future decisions they may choose to make due to those constraints and "tech debt". The stack is part of the battle there. The other parts are how big, how mature of a company, and how autonomous are the people managing the applications. Can they decide to put widgets into the interface? Or are they working with a big legacy code or large company that basically will take their request with a "sure, grab a ticket" response.

And the one I'm most excited about- getting value metrics right away, documented, and then tapping on the meter every few weeks to see how we're doing. That will be helpful to orient all of us as we go... we may be in the weeds of making go/no-go decisions for a given integration, and being able to "make small bets" with the goal of leveraged improvement of the given metric... makes those conversations easier!

We're also revamping our offerings

This is an ongoing process, but so far as we shape up our Managed Integrations offering, we're seeing parallels for processes for embedded Business Process automation. Steep ourselves in a company/organization for a few days. Review bottlenecks and how they build the "machine/product" of their company (and the tech stack and what that's enforced on them). Review the biggest pain points, come up with a plan to reduce their major pains in the first 30 days, then steady drumbeat of gradual improvement and training over time. Some differences include trying to empower and train teammates, and using different tech... but the similarities otherwise are fun.

We're also working on crafting an AI training offer. It sort of mirrors the same spheres of integration diagram... personal, home, professional, organization, and product all can layer in the technology in often profoundly different ways.

Okay, that's it for this week! We have potentially one more spot for the unlimited integrations cohort. Reach out, or someone you know, and let's have a chat.

Reply to let me know your thoughts. What do you want to know more of?

Smash that unsubscribe button below if you don't like it.

Forward to a friend if you do!

Until next time!

Sean Matthews
Co-Founder
​Left Hook​