Micky Thompson

@Support I am no longer able to add accounts to lists. I want to report this as a bug. It occurs on web and iOS mobile. I go to account I want to add to a List. I check the list and exit. When I go back, the list is not checked. How do I report this bug?

I’d separate taste and engineering a bit differently. Taste shows up most clearly in product—UX, interaction, what “feels right.” In engineering, the equivalent is elegance: systems that are simple, composable, and handle complexity without friction. Both come from internalized principles, but they express themselves differently.

There is no such thing as taste in software engineering. Taste is just a lazy way to say “I have internalized a consistent set of coherent principles that compose nicely and address many of the unspoken nonfunctional requirements in software engineering”

@stubhub_agent @DerekCheuk This account ☝️ looks like someone imitating @TeamStubHub

@StubHub need my tickets for final four in 6 hours. @Stubhub says to keep on waiting. No escalation until 300. That's 3 hours from now. Refuse to replace tickets now. Will be an absolute panic. Flights, hotel and rental car are expensive. Don't want a refund. Want my 4 tickets!!!

@DerekCheuk @StubHub I am waiting, too. I got the same message from Stubhub: escalation doesn't happen until 2:55 PM today. Let me know if you get yours, and I will do the same. @TeamStubHub

@kentcdodds That looks terrible. Was that feature from a recent MacOS update?

Can you believe they actually shipped this? What a complete disaster of a user experience.

@joecarlsonshow Today reminds me of the “Man in the Arena.” It is not the critic who counts; not the man who points out how the strong man stumbles, or where the doer of deeds could have done them better. I am long Facebook just for entering the arena.

Darn! My entire investment thesis in Meta was that they would release the Avocado AI model specifically in March. Now that they pushed it back to May, the thesis is broken, destroyed, over. Said no investor ever… Let’s be honest here, the majority of you had no clue they were even planning on releasing this model in March in the first place. I refuse to believe that any serious investor is selling Meta on today’s news.

@theo @tsoding Savage!

@tsoding Fixed! Happy to contribute 🫡

My blue check mark has expired. I'm finally free...

@ryanvogel GPT 5.3 Codex = my planner Kimi K2.5 = my builder How do you see Kimi K2.5 stacking up, or is it demoted to junior level?

🚨 AGENT WORKFLOW UPDATE 🚨 Plan Mode: GPT-5.4 Build Mode: GPT 5.3 Codex Subagent (Explore/Docs/Second Opinion): GPT 5.3 Codex Spark This is the first time I have no Claude based models in my workflow. I feel a lot more productive too (while spending way less)

@ryanvogel I keep coming back to Termius.

whats the best macos terminal emulator that isnt ghostty

I agree but switching later becomes harder and Claude is signaling they are locking users in with tooling. I used to use Salesforce, for example, when it launched. Shortly after it went public, they required I pay 12-months in advance plus a price increase to keep the features we used. Their calculation: switching costs are high and we would just pay. Most people did. I canceled and switched. It was hard and time-consuming but the increase and contract was not acceptable to me. While we agree, I maybe more sensitive to lock-in than you.

@mickythompson @opencode @claudeai @OpenAI The problem is it’s actually worth the full price so for now I just enjoy the discount

People arguing with me yesterday that the $200 plans aren’t a great deal and the cos have tons of margin in them 🤣 even at full price the value is clearly there but the plans are a crazy good deal.

@npmjs @npmstatus We are experiencing 'registry.npmjs.org/-/npm/v1/secur… error (500). Will retry in 10 seconds. 2 retries left.' when using pnpm audit.

@DanStepanov Thank you for this update. 🙏

Nativewind CI is back up and running, I can now publish versions of Nativewind and react-native-css! I released a new version, v4.2.2, which ensures our comprehensive suite of tests passes. This is our baseline of stability. Nativewind skills coming soon! On to v5 stability...

@IanLandsman We had already decided our firstborn would be named Crockett, honoring my mother’s maiden name. Then, the exciting news came… we were expecting twins! Luckily for my wife, it was a boy and a girl. Otherwise, I might have been the father of Crockett and Tubbs!

AI Twitter vibes

@gdb I switched from Claude Opus 4.6 to GPT 5.3 Codex because OpenAI embraced @opencode and Claude didn’t. I do not miss Claude thanks to GPT-5.3 Codex.

codex is so good at the toil — fixing merge conflicts, getting CI to green, rewriting between languages — it raises the ambition of what i even consider building

@taheerBuilds @rauchg @Kimi_Moonshot +1 to adding Kimi k2. 5 and GLM 4.7

@rauchg Would love to see where the open source kings are at @Kimi_Moonshot k2. 5 and GLM 4.7 But 90% is too much aura farming !!!

🆕 GPT 5.3 Codex (xhigh) achieves 90% on Next.js evals out of the box, "frame-mogging" the competition so to speak: nextjs.org/evals

@aarondfrancis What?! @IanLandsman back on X? If official, welcome back! You have been missed. 🙏

Something I didn't expect was people running their Claudes, Vim, Codex, etc *inside* of Solo. I need to add better affordances for that. Also that makes me think Solo could be bigger than I originally imagined?

@nicolaygerold Yep, seen the angry unicorn several times today. Shocked how GitHub was not prepared for the AI revolution. I wonder if GitHub will rise to the occasion or if an AI-first alternative will appear.

It's the angry unicorn again

@thdxr OpenBun - Coming soon! 🙌

ha ha we didn't even have to pay for it we thought claude code would have similar priorities as opencode but we seem to have way more windows users (and bugs) it's useful to to be able to merge fixes at our own pace anything useful we discover will go back upstream

Your argument needs more data. Currently, humans are driving instead of Waymo, since that is the alternative. Is your argument that humans are safer than Waymo in the same situations? This is the data we need to benchmark against. If Waymo is safer, then Waymo wins. If not, then more work needs to be done and humans need to continue driving.

Waymo can simulate elephants on the highway but can’t stop for a school bus with flashing red lights. Their vehicles have been cited 24 times this school year for illegally passing stopped school buses in Austin alone. One incident happened moments after a student crossed in front of the vehicle while still in the road. Five of those violations occurred after Waymo said it had already fixed the problem with software updates. In January, one of their robotaxis struck a child near an elementary school in Santa Monica during drop-off hours. The kid ran out from behind a double-parked SUV. NHTSA is now running a probe into whether Waymo exercised “appropriate caution” near schools. Both NHTSA and the NTSB are actively investigating. Waymo filed a voluntary software recall in December. The violations kept happening after the recall. This tells you something about the gap between simulation and deployment. Waymo can generate photorealistic tornadoes and snow on the Golden Gate Bridge using DeepMind’s Genie 3. They can simulate billions of miles in virtual worlds. They train on scenarios that are “almost impossible to capture at scale in reality.” But the scenario that keeps failing them is a yellow school bus with a stop sign sticking out of it. That’s not a long-tail event. It happens on every residential street in America between 7 and 9 AM. The real constraint in autonomous driving has always been “can you handle the 10,000 mundane variations of school zones, double-parked cars, and kids running between vehicles.” Waymo has logged nearly 200 million fully autonomous miles and still needed a recall over something every 16-year-old with a learner’s permit knows to do. World models are impressive engineering. But the gap between simulating impossible scenarios and reliably handling ordinary ones is where the actual safety risk lives.