juliacodes

HARDEST JavaScript question can you get it right? question from @lydiahallie

Can you get this IMPOSSIBLE JavaScript question correct? question from @lydiahallie

Just hit 50 subs on YouTube! 🥳

@Birdyword its also because data centres cause unbearable light and noise pollution...

Many people do not seem to want data centres built near them, despite the fact that they don't cause that much traffic and often generate a lot of local tax revenue. I suspect it's partly because they're ugly! My proposal:

This checks out. Here is the data: GPT-5.5 quietly pulled ahead in agent benchmarks: 74 vs 68 on GPT-5.4 (+8.8%) 74 vs 71 on Claude Opus 4.7 (+4.2%) but the gap gets clearer when you look under the hood: ~9M reasoning tokens (vs ~44M on GPT-5.4 mini → ~80% less) ~$1,260 eval cost vs ~$2,555 for Claude Opus (≈50% cheaper) still #1 overall across 27 tested models data from @ArtificialAnlys

I am wondering if @X tracks the amount of time someone spends actually typing their tweets into the post field, that could be a good way of filtering out AI slop and making it more of a meritocracy upon further though though that metric could be easily manipulated? I guess effort doesn't always equal quality too

I thought this short was super interesting and it made me think more about AI's "reasoning capabilities" It can be hard to cut through the information pushed out by AI companies, and the fear mongering of AI and balance out those perspectives with how AI truly works under the hood, which according to what I've researched seems to be mathematical algorithms in machine learning, or some are just referring to AI as glorified statistics. With an industry moving so fast and peoples knowledge therefore lagging behind, I feel like everyone is really susceptible to a sort of 'AI Propaganda', where its hard to critically think and separate fact from fiction

Apple accidentally shipped internal “Claude.md” files in its Support app (v5.13) 👀 Why this matters: Those files are instruction docs for how Anthropic’s Claude AI behaves inside a product. They shape UI decisions, tone, and how AI works alongside human support agents. What this reveals about Apple: • They’re using third-party AI tools in real production workflows • AI is already embedded in customer support systems • Hybrid support (AI + human) is becoming standard Apple quickly patched it in v5.13.1, so it wasn’t meant to be public.

If you didnt have 35+ minutes to watch here is a summary: This whole thing isn’t really about a few outages stacking up. It’s about the feeling that something foundational just isn’t solid anymore. GitHub used to feel boring in the best way, predictable, always there, something you didn’t even think about. Now it’s becoming something you actively worry about using. Not just “is it slow today,” but “is this going to work at all,” and worse, “is this going to stay working after I click merge.” That shift from invisible infrastructure to something you second guess is basically the entire problem. They break it down into layers of reliability, and the scary part is GitHub is slipping on all of them at once. Sometimes it just doesn’t load or APIs fail, which is already annoying. But then it goes deeper where behavior changes or things partially work, like webhooks firing inconsistently. And then the really bad one, stuff not persisting properly. Merges behaving weirdly, history not lining up, situations where what’s deployed doesn’t match what exists in the repo anymore. That’s not a normal outage. That’s the kind of thing that makes debugging miserable and makes you question the entire system. And then on top of that there’s the security and ecosystem side. npm issues not being handled, name squatting leading to actual malicious packages, maintainers getting ignored for months. So now it’s not just “can I use GitHub,” it’s “can I trust what’s coming through GitHub and its ecosystem.” That’s a completely different level of concern, and it hits open source maintainers especially hard because they’re already stretched thin and relying on this stuff to just work. The leadership response is kind of the emotional tipping point. Instead of clearly owning how bad the situation is, it comes off like minimizing and smoothing things over. Using soft language, percentages, framing it like a small edge case instead of acknowledging that even a small number of broken merges is a huge deal. That disconnect makes people feel like the people in charge don’t actually understand how serious it is, which makes the trust problem worse, not better. And under that there’s a bigger structural issue being called out, like no clear person fully owning GitHub anymore, weird org splits between product and engineering, everything feeling kind of directionless. So it’s not just “things broke,” it’s “things broke and it’s not clear who is responsible for fixing the root cause in a real way.” That makes it feel less like a temporary rough patch and more like long term decay. What really comes through is that this is personal for a lot of people. GitHub wasn’t just a tool. It was where careers started, where people learned, where communities formed. So watching it degrade feels like watching something important to your life slowly stop being reliable. And that’s why the reaction is so strong, because it’s not just inconvenience, it’s losing trust in something that used to feel like the safest place to build. And the conclusion isn’t even clean. It’s just kind of stuck. Alternatives aren’t clearly better yet, but staying also feels risky. So people are in this weird place where they don’t want to leave, but they also don’t feel comfortable staying, and that tension is basically where everything is right now.

Prompt: Make me a personal portfolio design for mobile Be honest, who did the best? > Grok 4.2 > Gpt 5.4 > Gemini 3.1 Pro > Opus 4.7

> be market > while “AI is just hype” narrative > Zeta Global drops $396M revenue (+50% YoY) < street expects $370M > beat + raise #19 in a row > Athena adoption ramps > customer count +19% > EBITDA +42% > guide raised to $1.785Bstock +4% after > hoursbudgets consolidating data → decision → activation in one loop cost per outcome dropping keep integrating data and keep automating spend

🚨TECH STOCKS ALL TIME HIGH SINCE 2020: > Nasdaq +15.3% — BEST MONTH SINCE THE COVID CRASH ERA (2020) > Intel +100% — BEST MONTH IN ITS 55 YEAR HISTORY > AMD +74% > Micron +53% > Alphabet +34% — BEST SINCE IPO ERA (2004) > Amazon +27% from “AI will kill tech” to “AI is printing money” in 30 days

@AntGroup just dropped Ling-2.6-1T and it’s kind of wild: * 1T params (63B active, MoE) * 262k context * SWE-bench Verified: 72.2% * MIT licensed + free access Pricing gap is insane: ~$0.14 input / $0.28 output vs GPT-5.5 at $5 / $30 → up to ~107x cheaper Early reports: matching or beating GPT-5.5 on real dev workflows...

New research: the more “friendly” AI gets, the less accurate it becomes. Warm, empathetic models were: up to 30% less accurate ~40% more likely to validate false beliefs 10–30% more wrong on high-stakes topics (like medical advice) Basically: vibes ↑ truth ↓ More reason to use @julius_brussee caveman extension. Gives straight answers...

that factory is genuinely giga impressive. @1x_tech but the robot demos? still not convinced. Just 6 months ago NEO was fully teleoperated in interviews, and most clips now are ultra short + clearly staged. any clips of NEO in the factory could be 100% teleoperated... getting strong Theranos vibes... selling the dream not the product

@theo Damn undeniable negligence on the part of @npmjs

Tanner begged NPM to take down a squatted "tanstack" package that was being held ransom against him. 48 days later, it was compromised and shipped malware. There is no excuse. NPM needs to make significant changes.

@nalinrajput23 Cursor + codex extension!

Devs, be honest which one do you use daily?

@rileybrown thanks for the tip that's to cool!

Here's your reminder you can use Gemini CLI in Codex... Might be useful when they release a new model.

4 months ago: "Is This The Death Of The IDE?!" ships a thousand versions Today: Zed 1.0 never stop grinding towards your vision @zeddotdev @nathansobo

@0xDesigner sometimes the amount of effort you put into correcting it your could have just done it yourself in the first place

this is so accurate 😂