--- type: "article" title: "What’s Cookin’ in Hook-land? (Vol.3)" summary: "Vol 3, on news, businesses facing scarcity, and a new offer from Left Hook" newsletter: "Left Hook" newsletter_handle: "left-hook" newsletter_url: "https://usecommune.com/n/left-hook" author: "Sean Matthews (@seanspeaks)" published: "2025-07-13T19:05:41.000Z" canonical_url: "https://usecommune.com/n/left-hook/a/what-s-cookin-in-hook-land-vol-3" markdown_url: "https://usecommune.com/n/left-hook/a/what-s-cookin-in-hook-land-vol-3.md" chat_url: "https://usecommune.com/n/left-hook/a/what-s-cookin-in-hook-land-vol-3/chat" source_url: "https://lefthook.kit.com/posts/3" body_source: "imported" likes: 0 replies: 0 body_words: 2623 --- # What’s Cookin’ in Hook-land? (Vol.3) *Here's the bet... that I can keep this up for a year. Let's hope this doesn't go the way of my work-out routine. Although to be fair, I love talking and writing. I do not love working out. ​ I added everyone I thought might be interested in a quick and thought provoking read… please don’t hesitate to unsubscribe! And on the flip side, please reply and let's talk about whatever catches your attention. Okay cool. Let’s rock.* ***(Oh and special note, at the end, we're launching a new experimental offer for software product companies... unlimited integrations, fixed monthly fee.)*** # On a personal note... This personal note is an easy one. Next week I'll be going on a vacation. To Ocean Park, Maine. My mom is one of 8 children (my dad is one of 6). (... Yes, Irish Catholic 😂). Since she was around 6 or 7, they've been going to Ocean Park as a family for at least one or two weeks every summer. That streak has remained unbeaten for 60+ years now. So, we continue it :-) ​ The hard thing to know is how long it will last. When it was 10 of them, they just needed one cottage. We've managed to rent about 4-6 cottages to the whole big family every year now for the past 30+ years as they've almost all had kids. But now... my family is the oldest (while I have about 30+ first cousins total, none of them are actually my age, and on my mom's side, myself and my two older siblings are older than everyone), and we're starting to need 2-3 cottages. The other families are bound to get bigger as the cousins get married and face their own family pressures. We're blessed with a pretty functional family that enjoy gathering every year, so it hasn't exactly been pulling teeth to get everyone to keep coming back. But they all have to travel... for me it's 45 minutes. For some of them is 12 hours or more of travel time. That's $$ and coordination. I'm just waiting for the moment where someone consistently stops coming. And then waiting to hear how everyone handles it. ​ I know we look forward to it every year, and now that the kids are starting to be of the age where we don't have to worry about them eating sand or getting into trouble with the waves and undertow, it's actually relaxing to go to the beach with them. (Side note- I really do not enjoy baking in the sun. Game changer was when we got a pop up tent. I can do that shade all day.) ​ So... expect to get *something* next week. I haven't fully committed to a full week's write-up. But I think it will be fun to send you a photo of the beach, just like every photo of a beach you've ever seen. Maybe I'll throw some integrations in there, somehow. ​ The crossing jet lines into a cloud as a hub-and-spoke model diagram? ​ Some story about "integrating" our "young family" life schedule into our older family's lives with their late wake-ups and arriving at the beach just when naps are starting? Who knows. ​ I *AM*having the monthly business planning review with my family members (retired mother who worked at Fidelity for 25+ years, my wife, and my soon to be brother-in-law who works at McKinsey) while on the vacation. ​ But trying to stay unplugged as much as possible. Wish me luck! ~Sean # What we’re seeing **AI Browsers? Aren't those just browsers?** Okay, if you've missed it, we suddenly had the release of a number of "AI Web Browsers", and I'm fascinated. Despite the fact that no model has proven to be truly intelligent and avoid the hallucination issues, we're plowing ahead as a species to layer in LLM-based AI everywhere. I'm game for the trying! To wit: - ​[Dia](https://www.diabrowser.com/)​ - ​[Comet](https://comet.perplexity.ai/)​ - ​[AI mode for Chrome](https://www.google.com/chrome/ai-innovations/)​ - ​[Brave has an Agent baked in](https://brave.com/ai/)​ - Whatever it is that Microsoft Edge and copilot is doing And I believe a few more. I do like the idea that the agent can use your user session, and you can make sure it asks you to do stuff before it does it. And it can take action on inactive tabs. Some cool use cases come out of that. On the product integration side of things, someone took the concept of MCP and applied to this paradigm: "[MCP-B](https://mcp-b.ai/)"... aka an MCP server that you can implement as a web-app for browser agents to leverage. What does that mean? It means that beyond developing your API to be consumable by LLM/chatbot tools (MCP), there's now a path to develop your existing web app user interface to be consumable by LLM/chatbot tools (the browser agents). In a statement- "User experience" designing just got way more complex. Even if this specific implementation or concept doesn't catch on, it's effectively what people have been trying to figure out with robots.txt or llm.txt (no standard, yet)... the difference is this is for live AI co-browsing. **Updates in the integration workflow space?** A few other things we saw this week that caught our attention... - ​[2.0 for Paragon](https://www.linkedin.com/posts/brandonfoo_as-of-today-paragon-is-no-longer-an-embedded-activity-7348027054736519169-gjWs?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAebMZ8BQwmRCusyaW979DrWyU07mcA4Vxo) "No longer an embedded iPaaS" strong claim! But also likely reflecting the way their product has been used by a subset of their customers. They still support their embedded iPaaS approach (Workflows) but have layered in a few other pieces that take them in a broader direction. ​ - ​[Notes from interation.app](https://www.linkedin.com/feed/update/urn:li:activity:7348339195477696513/) CEO Daniil Bratchenko His platform's architecture and proposed "Full Stack of building integrations in the AI era". The thing I really appreciate about their approach is that AI isn't required at every stage of the approach, but it helps spin together the glue and structure. ​ - ​[Cyclr CEO's note about AI in iPaaS](https://www.linkedin.com/posts/fraserdavidson_the-role-of-embedded-ipaas-in-the-ai-gold-activity-7324072476043264001-g6TJ?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAebMZ8BQwmRCusyaW979DrWyU07mcA4Vxo)​ Commentary around AI not belonging in the product for embedded iPaaS tooling. I've known Fraser for almost a decade now and he's been both insightful and consistent with his messaging and thinking about the use cases and needs for his tool. Good to scroll through other posts from them! ​ - ​[Arcade.dev co-founder's piece about MCP's and APIs](https://www.linkedin.com/posts/alexsalazar_the-mcp-ecosystem-just-hit-a-major-inflection-activity-7349567800379695104-PxP-?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAebMZ8BQwmRCusyaW979DrWyU07mcA4Vxo)​ This one is a reflection on two other recent pieces. What I appreciate about this was his ultimately point that- you need to treat your MCP servers/tools as another product, and NOT just autogenerate them. For those non-developers in the audience, basically this is a statement in support of the idea that there are going to be some really well designed and iterated tools for your AI chatbots to connect with... and some really poorly designed and unusable ones. ​ Anything you think I missed or is worth adding to the mix? # What we’re hearing I had a conversation with [Fraser Davidson](https://www.linkedin.com/in/fraserdavidson/)recently that included a few memorable one-liners. The one that really felt relevant this week: when there's a fire burning everywhere, integrations are one of the first things dropped and left behind by the team. Something to that effect, anyway. I chuckled because over the years we've definitely felt that. Especially when things get tight fiscally at a company. And, when it's firefighting, the person running point often can't take the time to document what was in flight or being worked on when the left. And then when the dust clears and someone new comes along to pick things back up, they're kind of in the lurch. **Product managing an integration takes time, and time is in short supply** That heading is basically the summary of most of our conversations this week. Whether from people leaving a company, or from being stretched thin, or from shifting priorities, I had conversations with 4 different companies where scarcity of some resource was a main concern. In some cases, the scarcity is in domain expertise or knowledge (like understanding how Sales Reps use HubSpot in larger companies), and the remedy is either to find someone with that expertise, or doing some digging and ideation, or listening to your users. In all cases, that's more time. So, what do you do in moments like that? I had a realization that one of the valuable things we can provide at Left Hook is the combination of longevity and broader ecosystem view. Our day to day is living in the spaces between products and sales cycles etc. And we're far from the only company or group of experts like that. Look for those experts and what they post about, write about, or resources they've put together. It can save you time (and therefore money). And obviously, ruthlessly prioritize. **Designing and launching integrations is still a slog** That goes hand in hand with another thing we're hearing. Not directly, mind you, but just purely in timelines being discussed. We've had integrations ready to launch in some cases for months, and are just now getting the GTM attention they deserve. That was known at the front of it, but it still boggles the mind that it took that much time. Or just in the process of designing an integration, it can take time purely to line up the right people, for the right calls, before committing getting "pen to paper" (code to repo). That time can probably be condensed if done right, which lead me to making the offer I'll note below... but it's still a problem that feels a bit intractable and somewhat untouchable via AI at the moment. At least until someone has trained an AI on years of expert data and keeps it up to date and relevant. (Wait that sounds familiar) ​ Anywho, let me know if either of those resonate with you, or if you've solved this for yourself. I'll take whatever tricks you can provide and share them broadly! # What we're doing Because I liked doing it this way last time, here's the overlapping automation diagram again: ​ ![](https://embed.filekitcdn.com/e/gMtzGmZ1nP1uogiQWhDnPA/5952zDDGCixHotopvkNZuw/email) *​* ​ So, what we're doing for each one: - **Home- ​**I had fun and switched up the Alexa skill so now every day of the week has a slightly different saying that gets announced at home. Scout (our 2 year old) saw me walk in and said "It no say 'don't forget to wash your hands, fam!'"? with sort've a question and excitement. My next trick is to create an Alexa skill that auto-generates a random one every day. I bet Claude code can help me do that :-) ​ - **Personal-**​ I've been really playing around with Reuven Cohen's work on "agentic swarms" for use with Claude Code. He hasn't talked about it a lot, but he created this [side project called "SAFLA"](https://github.com/ruvnet/SAFLA) (Self-Aware Feedback Loop Algorithm) that I've forked and been trying to layer into my coding work. ​ I've posted about this on LinkedIn and will have other posts, but the fun thing: I now have a shared memory that my "claude code" instance can tap into across context across repos/projects. ​ And I added in a bookmarking system. So after I've done something in one project, I can tell claude to "bookmark this", and then hop to another project, ask it to bring up the bookmark and implement a similar thing in the current project. ​ It's been much nicer for context switching and I really notice whenever I go back to non-memory AI prompting. Feels like I'm rebuilding that context from scratch every time. No bueno. ​ I'm working on making this a shareable memory tool so that Daniel and I can share patterns and contexts, synced to a cloud DB in the background. ​ ​And at some point, there's no reason it can't be plugged in to a number of other contexts and flows fully outside of development ​ What I'm really looking forward to is patching this into any LLM system I'm using across the business and personal life, and basically having my own personal and business memory system following me around. ​ Feel free to ping me if that sounds interesting, we can collaborate. ​ - **Work- ​**I didn't do too much on the automation front for our day to day work this week. I did have a revelation that this newsletter will be faster to write if I just chipped away at it as stuff happens during the week. And the automation piece is to aggregate the notes into a pre-build newsletter so that I'm playing the role of editor moreso than writer plus editor. ​ The one tip here... I think ChatGPT is better with spreadsheets in a PDF format, at least for comprehension (row by row or cell by cell manipulation aside). ​ - **Organization-**​ We've been working on our improvements for our support and maintenance contracts and monthly emails that we send to folks. This automation set runs through Zapier, uses GitHub commits and PRs, summarizes them, and generates HTML for the email, alongside our own notes and stats we pull from various sources. ​ I appreciate having someone else to collaborate with in these moments, since I had a hard time committing to one of the platforms or just rolling our own... the curse of knowing how to do the same workflow in any one of a dozen different tools plus pure code. ​ So that's a tip. If you're getting some paralysis by analysis, rope someone else in briefly, have them make a gut choice, and get cooking! ​ - **Product- ​**One of the pain points for folks getting Frigg to production is getting the integration deployed into their AWS accounts, with the right permissions, and with a certain level of security. ​ We've been ironing out the kinks this week on the permissioning side and added to the forthcoming **frigg cli**... so now when developing with frigg, once you've got your app definition set, you can just prompt \`frigg generate credentials\` and it will prompt you to select the cloud provider, the infrastructre-as-code tooling type, and spit out a file for you to patch into your existing workflow. ​ Or if not "you", at least handing it to DevOps gets you much further faster. ​ 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​ ​ **PS- NEW OFFER ALERT** So, I teased this at the top, but throwing it down here... we're trying out a new experimental offer for the first 5 companies that take us up on it. $5,000/month, unlimited integrations, guaranteed 10 launched in the first month. A few things conspire to make this even possible but also desirable for me. First, we're so much faster now at building the integrations that the bottleneck is no longer the code that powers them, but the product management and go to market around them. Second, we've seen and built so many integrations that at this point we find ourselves pulling people along in a design or go to market process, when we'd be much faster and more iterative if we could just be given the reins and told "go". Third, we know the value of integrations for you and your end users is NOT in them existing. It's in the outcomes they provide, over time. And the earlier in market you are, the more time horizon you give yourself to realize those outcomes. ​ I've been trying to figure out the right model for years to tap into this potential. This feels like a potential winner. ​[Here's the doc outlining the offer and approach.](https://learn.lefthook.com/unlimited-integrations)​ ​ Respond if you're interested. Let's jump on a call. ​ *** ## Discussion No replies yet.