TJ @tjtjtj.tj · 02/10/2026this is how they described it in the email I got for the DRLE photoshoot 240
TJ @tjtjtj.tj · 28/09/2026The one on the left is fake, confirmed via SynthID watermark from OpenAI. Right one seems real and corroborated by news reports. 120
TJ @tjtjtj.tj · 17/09/2026If you’re shopping for conjunctions don’t bother going to Safeway. They don’t even have “yet” or “so” 121
TJ @tjtjtj.tj · 06/09/2026reminds me of this guy who tried defending cybercabs with a similar argument... like, yeah, that theoretical guy he's referring to is *actually right* lol 010
TJ @tjtjtj.tj · 05/09/2026these old-Aperture posters from Portal 2 are particularly prescient these days 070
TJ @tjtjtj.tj · 20/08/2026it hasn't really changed since 2022 though. if I saw this for the first time today I would definitely say this is vibecoded, but it's been this way for a while so I don't want to sully the designers'/developers' honor by claiming so 210
TJ @tjtjtj.tj · 20/08/2026this thing they added since is definitely claudepilled tho lol. surprised it doesn't use a • as separator 110
TJ @tjtjtj.tj · 15/08/2026I see what you’re up to @transit.app !! Been waiting a long time for this exact feature 070
TJ @tjtjtj.tj · 10/08/2026example: the sign in the center says "Scooter and Bike Share Parking" - could just move it a bit to the right and demarcate that space 010
TJ @tjtjtj.tj · 05/08/2026Absolutely not compression artifacts, that wouldn’t explain the different pixel sizes. As I mentioned, the images have a SynthID watermark from OpenAI, meaning they are certainly AI-generated. (Unlike other “AI detection tools” which basically just guess, SynthID watermarks are deterministic.) 000
TJ @tjtjtj.tj · 05/08/2026It still does, just zoom in a bit. No coherent/consistent pixel sizes and a splotchy mess. 110
TJ @tjtjtj.tj · 04/08/2026and the lack of attention to detail bugs me too... why are the 1 and 2 Line colors/shapes like this? any transit nerd would have immediately caught this. 130
TJ @tjtjtj.tj · 24/07/2026my solution is to make agents snitch on themselves; I put this (1st screenshot) in my PR template, tell them to follow it exactly in AGENTS/CLAUDE[.]md, and it's produced good results so far (2nd/3rd screenshots). the goal is to better understand the roles of the agent and human involvement 250
TJ @tjtjtj.tj · 20/07/2026this is an AI-generated map; the information doesn't look incorrect, but it's better to link the actual source (I don't know why whoever generated this couldn't just take a screenshot?): mapresearch.org/equality/ 2484
TJ @tjtjtj.tj · 03/07/2026this is incredibly strange, my (Chromium-based) browser apparently has a built-in style to hide `.scad` elements, so the name isn't appearing in the table. what the hell 100
TJ @tjtjtj.tj · 11/06/2026woah, King County Metro is reporting occupancy in their GTFS-RT data now. I can't speak to its accuracy yet, but it's cool they added it 3112
TJ @tjtjtj.tj · 04/06/2026Even a small reduction in speed dramatically improves outcomes for collisions. Unless you enjoy traffic violence, drive the speed limit. www.ite.org/technical-re... (Also: this is in Bellevue, WA, USA. Not sure where you got Canada from.) 200
TJ @tjtjtj.tj · 26/05/2026Write out a quick bodge and we have something that parses the archive into something usable. 160
TJ @tjtjtj.tj · 26/05/2026And we can get the route color by looking at the background color metadata (which is helpfully in a JSON blob near the end of the archive). Same deal: look for this pattern, extract and convert the color values. 160
TJ @tjtjtj.tj · 26/05/2026This format is proprietary and would take a fair bit of effort to fully reverse-engineer (digging into private frameworks and such), but there is a shortcut we can take: the accessibility descriptions are baked in. All we need to do is look for this specific pattern, no complex parsing needed. 170
TJ @tjtjtj.tj · 26/05/2026So how does it actually work? Transit does not offer any official interface for this sort of thing, and it's basically impossible without a jailbroken iPhone to read data from third-party apps. Recent versions of macOS include a Continuity feature that mirrors iPhone Live Activities, shown here. 160
TJ @tjtjtj.tj · 26/05/2026I love using @transit.app's GO Live Activity to keep tabs on when I should leave, but I always wished it was more glanceable while my phone is on the charger and I'm scrambling to get ready. The solution: make it mirror to my Transit Tracker! 3539
TJ @tjtjtj.tj · 08/05/2026fun fact: all the calculations happen client-side, there is no routing server! this was a fun challenge to implement and optimize. at build time it generates a bunch of GeoJSON tiles representing the transportation network, then the client pulls only the tiles it needs to build a graph around stops 130
TJ @tjtjtj.tj · 08/05/2026it now calculates walk-time isochrones instead of a static radius around each stop - you can change it based on how long you're willing to travel from the destination stop 160
TJ @tjtjtj.tj · 07/05/2026if you live in the Seattle metro area and want to find a reason to take new bus routes then boy have I built the tool just for you: go-by-bus.horner.tj (featuring @transit.app's web planner) 03114
TJ @tjtjtj.tj · 25/02/2026it also says Feb 25 for me in the US. either the email or text here is wrong. I'd have to assume they intended to push this change out by today but the timeline got pushed back and they forgot to change the date 100
![Screenshot from the text view of a hex editor, with the text highlighted: "{"backgroundTintColor":[0.6354607,-0.061623164,0.14334208,1]}"](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:bmkraaatxmnlb3ytxhnxbog3/bafkreibk5ymvv2p4wqgq5n5bpluytzios2cixwv45aymoxz2bf34p54whu)
![[check box] Share my shipping address with sellers
This setting will no longer be supported starting February 25, 2026. After this date, third-party sellers will receive your shipping address to fulfill orders. You can review or update your list's shipping address on this page.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:bmkraaatxmnlb3ytxhnxbog3/bafkreigfily7t6t2izk4hbow2hqqdzwnrv7ewteue6yicym2fftkismvom)
!["I spend a lot of time on this task. I should write a program automating it!"
[Two graphs are shown, plotting workload against time.]
Theory:
[The line for "work on original task" is steady but then drops down to a much lower level.]
[The line for the automating job increases heavily while "writing code" and then drops down when "automation takes over".]
[Both lines end up with a big amount of "free time".]
Reality:
[The line for "work on original task" is steady with no drop to a lower level.]
[The line for the automating job increases heavily while "writing code", then it increases again while "debugging", it drops down slightly while "rethinking", and grows up again with an infinite end while the task is still an "ongoing development".]
[The line for "work on original task" ends up with "no time for original task anymore".]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:bmkraaatxmnlb3ytxhnxbog3/bafkreihl6axrh6psoimm7zfzkhlkqbrghzgqv25i7h4uluimb5cistu6xy)
![map of united states with speech bubble pointing to every individual state which says "all the people from [other state] have moved into [my state]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:bmkraaatxmnlb3ytxhnxbog3/bafkreigbvc6hsfc7hn7suilytqm5ujnxdeifjwdqhgfzwefwia6whwrghm)
