Joe Fabisevich

@mergesort.me

Director of AI Product Engineering @ Pursuit.org. I used to make Twitter (better), then that all went to hell. Now I make puns, @plinky.app, and build.ms. Born and raised New Yorker, trying to do a little good and be the friend you made along the way. 🍕🐱⚾️

I don't want to install another keyboard to get better dictation at the cost of messing up the keyboard layout and making emoji harder to reach, I want to replace this button with my preferred dictation engine. 😤

The iOS keyboard with an arrow pointing to the system dictation button

What subscriptions do you have to stay aware of tech industry news and trends? I've been a Stratechery subscriber for years, read Lenny's Newsletter last year, and have been intrigued by Pragmatic Programmer. But I want your suggestions, particularly ones that would help non-technical people.

Today's agentic systems are such a mind-blowing unlock for people who know what they're doing. I'm currently not quite vibecoding a code review tool (I'm guiding it heavily after every large batch of work) and I'm able to achieve 500%+ performance improvements simply by giving Codex a profiler.

Milestone 7's implementation and
performance work is complete; the model
comparison remains open.
Key results:
• Decomposition: 500 → 9 git apply
processes; 8.59s → 1.30s in the
comparable run.
• Ul interaction: main-thread hangs 2 →
0; SwiftUl update time reduced 79%.
• Fixed scrolling: CPU time reduced 59%;
hitch behavior remained effectively
neutral.
• Loaded memory: 650,576 KiB →
148,592 KiB, a 77% reduction.
• Search still reveals the final atom, and
repository fingerprints prove exact non-
mutation.
• Complete validation passed all 240
tests plus warning-free CLI and clean
app builds.
• Post-work review found no
actionable issues.
The implementation and evidence are
recorded in PLANS.md 7. The work remains
unstaged and uncommitted.

Two years ago Maiori figured out how to use our old dry feeder so she could get more food when we weren't looking. She's figured this one out too but the button is capacitive and won't respond to fur, so we often catch her pressing her paw on the feeder trying to get food to come out. #caturday

Every day the cover of the NY Post says something like "Mayor Mamdani tries to set the city on fire but settles for shitting in the streets", while I walk out of my house to see no fires and no shit.

I've become *that* person. I queued up batches of work to deliver a few prototypes and features (with meaningful results), ran a Codex Security scan, and timed it to when my reset was set to expire so I could maximize the work done for my allotted tokens. And honestly, it kind of ruled.

Reset Codex
Friday, Jul 31, 2026
at 10:45 AM
EVENT TIME ZONE
New York (EDT)
Show Time Zones M

Plinky is proud to support @matt1corey.bsky.social’s #IndieAppSales. Go check out Plinky at it’s lowest price to date, and support an indie dev (me!) while you’re at it. ☺️ 👇🏻 Tap here on iOS/macOS to find a new low, low price for the next few days, with 50% off a yearly subscription to Plinky Pro.

Plinky – Easily Save Links | Enjoy Them Later

Do you save links but never revisit them? Plinky gives you one-tap saving from any app or browser, with folders + tags + search, reminders so you actually revisit what matters, and full customisation ...

plinky.app

I see so many posts on Mastodon like this claiming AI cannot do things that we have empirical evidence of it doing, because the harshest critics of AI continue to equate agentic systems to LLMs. You don't ask an LLM to write code, a coding agent is a complex system with many parts *around* an LLM.

ॐ0
Ruth - of systems & design
@RuthMalan@mastodon.social
"This paper [Naur's Programming as Theory
Building is so timely in times of LLMs
because it gives us a definition for
programming that an Al can never do, by
definition. A large language model can
never form a theory about what it is doing,
that simply is not a goal of a language
model. An LLM can not (consistently and
correctly) answer questions about the
code."
Jul 26, 2026 at 09:55 AM •

Hey, you. Did you know Plinky is on sale? Right now, for the next week. Just open Plinky and instantly save 50% on one year of Plinky Pro. I’ve recently dropped prices in over 100 countries, so at this point I’m practically paying you to use Plinky! 👉🏻 Tap here on iOS/macOS to redeem: plinky.app/pro

Everybody's heard that an ant can carry 50 times its weight, but not everyone is generous enough to sacrifice a Dorito to validate the hypothesis.

An ant on a picnic blanket carrying a burrito

I honestly forgot Charlie and I had recorded this until I saw it show up in my podcast player today because it felt more like a conversation with my friend than anything formal. If you want to hear about my college shenanigans, my launch story for Plinky, with minimal doses of AI, give it a listen!

Launched@launchedfm.com · 3w ago

"We made $5,000 in the first two hours. Definitely reach out to press, you idiot." 😂 @mergesort.me built one of the first ChatGPT apps — in 19 days. One email to John Gruber, a post on Daring Fireball, and RevenueCat alerts started going off in a bar. 🎙️ New episode links in the 1st comment below 👇

New Launched! 🚀 Had so much fun hanging out with @mergesort.me talking indie app development and, of course, thoughts on building with AI 😁

Launched@launchedfm.com · 3w ago

"We made $5,000 in the first two hours. Definitely reach out to press, you idiot." 😂 @mergesort.me built one of the first ChatGPT apps — in 19 days. One email to John Gruber, a post on Daring Fireball, and RevenueCat alerts started going off in a bar. 🎙️ New episode links in the 1st comment below 👇

"We made $5,000 in the first two hours. Definitely reach out to press, you idiot." 😂 @mergesort.me built one of the first ChatGPT apps — in 19 days. One email to John Gruber, a post on Daring Fireball, and RevenueCat alerts started going off in a bar. 🎙️ New episode links in the 1st comment below 👇