Listening to this, I suspect MCP & CLI are much of a muchness, but CLI has the short term advantage of opening up access to existing tools & authentications at the client, and today's skills have reduced token usage by replacing MCP discovery/ negotiation with "do this via that". Time will equalize.
I use AI a lot, but here I was being pushed by it - not leading it - on certain features. They were its not mine. My edge is to move fast, but no faster than I can retain my products approach & feels. I want consistency more than quantity.
Not to mention how easily I can generate markdown content from my code, and turn things like internal tables and bugly spreadsheets into simple (easily parsable/cleanly structure) HTML - a programmatic SEO (pSEO) win.
When I picked Hugo (gohugo.io) for my website it was 'cos it was a static site generator (in Go) and my new SaaS was Go. I had no idea when I was setting myself up for markdown in my editor that AI would come along and feast on all my content, understand it all, and be a powerful writing assistant.
Would you have it any other way? All the more engaging!
A startup needs to cut corners - or, frankly, they'd be just like the legacy incumbents - and never launch. The art is picking what you allow to be broken. Customer development helps you pick, but go pick and be ok with the things you allow to be broken, as you cajoling your market towards success.
Sure, I'm using AI as thought partner, as code monkey, as tester, as researcher, as transcript analyzer, and much more ... but I cannot exceed the speed of my own comprehension/ nurturing/ attention. I could feature slop, but where is the *service* in that?
(2/2)
"The enemy of art is the absence of limitations." — Orson Welles. In a world of unlimited agentic AI development, I'm choosing slower and steadier feature development for my SaaS.
(1/2)
Each week I mute Slack until next week. Each week, when it un-mutes, I remember why I like it muted. One day I'll delete it, but this is working for me. :-)
Yup, nobody spends as much time working/worrying/pondering the problem as you for your SaaS. Feed off customer enthusiasm, but bring your solution.
Some things we control, some we don't. :-)
It'll likely keep them in better condition long-term, too. Just don't let out of sight mean out of mind.
I was once told I was "the least Zen person they'd ever met" ... and I'm worse on caffeine. :-) That said, I still think you need to let people see your thoughts / plans / designs. You can resist the pressure to release, but enjoy some feedback. Maybe some screencasts to build in public / market. 😀
Despite not having a podcast, I want Outro shipping! 😁 @ianlandsman.com how about you let folks (@aaronfrancis.com) use Outro, with the agreement you *will wipe the database* every week, so no migration woes. You are missing early feedback & risking getting friendly with fixable mistakes. Ship it! 😁
I wonder when we'll get "AI/human pair programming" where reviewers chat with the AI over the PR. Now why did you do this? What is this change about? What *were* you thinking here? :-)
In the EMS communities I hear talk of triaging medical calls for Paramedic (advanced) or EMT (basic) to reduce Paramedic burnout, support EMT skill improvement, etc. Maybe you have product triage PRs for impact / value, then an engineer triage PRs for risk / complexity, to then target who gets what.
I'm finding more and more of my "normal non-techie" customers are phone/SMS based, and email is secondary, or far worse. Can you send them SMS? Maybe your coming mobile app will help. I realized I am email-centric, but they are increasingly NOT.
To get back to your original concern, I feel that reviewing PRs only feels overwhelming if they come fast/furious/unending, without an understanding of the value of each. Tell me "contracts w/o pipelines" will improve scads of customer experiences massively, and I'll happily review the PR in detail.
I liked how in your video you pointed out "I know you (customers) don't love change" so factor in your customers too. Less Annoying no doubts includes stability, and the copious non-code work to roll out updates that your customers can process. Focus AI on work there, less on slinging pull requests.
*Sustainably* more is better. As you say, you'll listen to your team and factor in their wear and tear.
For me, I think that "can do more 'cos AI" needs to become "ruthless recommitment to prioritization" to avoid uncontrolled feature slop. Simply do what we do better, not more. More unit tests, more automated tests, and deeper investment in a few more things, not a lot more. Don't let AI wag the dog.
AI is very "productive" ... I doubt that'll be the last sucky problem it produces. ;-)
Good to see what you achieved within the product from the video, but I'll be interested to hear what your team felt / experienced / learned from the week. Did AI-detox week follow? :-)
I appreciate when folks put into words/emotions what I'm experiencing, like Benedikt's ( @benediktdeicke.com ) conflicted "too easy to start", "too many balls in the air", "not finishing". AI gets you 80% quickly, but the last 10% / 20% is the real work. slowandsteadypodcast.com/episodes/in-...
While I'm tipping the hat, I'd not want to be without "Tab Search" either. Both make my chaos a little more navigable. ( With work * 2, personal and fire department profiles, I have lots of windows of tabs. Tab Search is my goto for ... "where was I working again?" & "Which tab had that open mail?")
In a sea of multiple tabs, in multiple windows, in multiple workspace, in multiple monitors ... I'm loving the "focused togetherness" of Chrome Split Views. I can review a document on the left, and extract a spreadsheet of ideas on the right, and they stay together (while my mind wanders elsewhere.)
That last line is what takes this message from 'interesting for writers' to 'unlocking for me', thanks! My experience is from a data point of one, which hardly seems well researched, but maybe it is enough to help other solo founders triangulate their thoughts.
I have a small pouch on my suspenders for spare medical gloves. It saves fishing around when somebody needs gloves in a hurry. My pockets are similarly stuffed, like yours. :-)
To be clear, I've added:
1) full features (e.g. admin notes, timestamps, etc. etc.)
2) "custom fields" - richer than tags (designed/considered.)
3) "tags" - lightweight / flexible / throwaway / user defined.
In today's world (even before AI) I feel that "generic" comes last. Code is table stakes.
My product is not a different category, but the product's special sauce is in the different approach to that category. "that + more". I needed to be known/discovered as a "Fire RMS" (which felt "somewhat stale") but then convert with the extras they didn't know they wanted, but then they loved.