Conor Wade

@joshu @paulg An old 930, some skill, and a B road is kind of perfect. Although I have heard some shocking stories about potholes in the UK recently.

@conorwade @paulg for driving around the british countryside?

I used to be slightly bummed that Jessica didn't care all that much about watches, but I've realized that not caring much means they all look the same to her, and that in turn means she doesn't always realize when I've bought a new one.

Opus 4.7 seems to respond before it arrives at an answer which results in some really bad answers.

@paulg @joshu I recommend closing the door on an air cooled Porsche 911. 🤣

@joshu My younger son predicted this, after noticing that I tended to collect things that used to be beautifully made but aren't anymore, and wondering what else was in that category. But I don't think it will happen. Cars are too big.

What’s old is new again. This reminds me of the API everything period of Web 2.0 in the mid 2000s. @jamespotter also pointed out that this time companies can’t bait and switch so easily by locking it up again.

This! The rubbish I keep hearing from AI companies actually takes away my excitement for this extremely important technology.

@jlongster Wish I was going to React Miami! Would love to play. I would also challenge @thdxr to a one on one game too. 😅

who's gonna be at React Miami next week!? I challenge you to a tennis match

@kenwheeler If this was a Black Mirror episode it would be panned as ridiculous and unbelievable. 😅

i actually couldn’t even make up something this bad. i’m speechless.

I miss the days when tech founders were just your typical megalomaniac lunatics. Much easier to predict their behavior. washingtonpost.com/technology/202…

🚨SHOCKING: Apple just proved that AI models cannot do math. Not advanced math. Grade school math. The kind a 10-year-old solves. And the way they proved it is devastating. Apple researchers took the most popular math benchmark in AI — GSM8K, a set of grade-school math problems — and made one change. They swapped the numbers. Same problem. Same logic. Same steps. Different numbers. Every model's performance dropped. Every single one. 25 state-of-the-art models tested. But that wasn't the real experiment. The real experiment broke everything. They added one sentence to a math problem. One sentence that is completely irrelevant to the answer. It has nothing to do with the math. A human would read it and ignore it instantly. Here's the actual example from the paper: "Oliver picks 44 kiwis on Friday. Then he picks 58 kiwis on Saturday. On Sunday, he picks double the number of kiwis he did on Friday, but five of them were a bit smaller than average. How many kiwis does Oliver have?" The correct answer is 190. The size of the kiwis has nothing to do with the count. A 10-year-old would ignore "five of them were a bit smaller" because it's obviously irrelevant. It doesn't change how many kiwis there are. But o1-mini, OpenAI's reasoning model, subtracted 5. It got 185. Llama did the same thing. Subtracted 5. Got 185. They didn't reason through the problem. They saw the number 5, saw a sentence that sounded like it mattered, and blindly turned it into a subtraction. The models do not understand what subtraction means. They see a pattern that looks like subtraction and apply it. That is all. Apple tested this across all models. They call the dataset "GSM-NoOp" — as in, the added clause is a no-operation. It does nothing. It changes nothing. The results are catastrophic. Phi-3-mini dropped over 65%. More than half of its "math ability" vanished from one irrelevant sentence. GPT-4o dropped from 94.9% to 63.1%. o1-mini dropped from 94.5% to 66.0%. o1-preview, OpenAI's most advanced reasoning model at the time, dropped from 92.7% to 77.4%. Even giving the models 8 examples of the exact same question beforehand, with the correct solution shown each time, barely helped. The models still fell for the irrelevant clause. This means it's not a prompting problem. It's not a context problem. It's structural. The Apple researchers also found that models convert words into math operations without understanding what those words mean. They see the word "discount" and multiply. They see a number near the word "smaller" and subtract. Regardless of whether it makes any sense. The paper's exact words: "current LLMs are not capable of genuine logical reasoning; instead, they attempt to replicate the reasoning steps observed in their training data." And: "LLMs likely perform a form of probabilistic pattern-matching and searching to find closest seen data during training without proper understanding of concepts." They also tested what happens when you increase the number of steps in a problem. Performance didn't just decrease. The rate of decrease accelerated. Adding two extra clauses to a problem dropped Gemma2-9b from 84.4% to 41.8%. Phi-3.5-mini from 87.6% to 44.8%. The more thinking required, the more the models collapse. A real reasoner would slow down and work through it. These models don't slow down. They pattern-match. And when the pattern becomes complex enough, they crash. This paper was published at ICLR 2025, one of the most prestigious AI conferences in the world. You are using AI to help you make financial decisions. To check legal documents. To solve problems at work. To help your children with homework. And Apple just proved that the AI is not thinking about any of it. It is pattern matching. And the moment something unexpected shows up in your question, it breaks. It does not tell you it broke. It just quietly gives you the wrong answer with full confidence.

This era of people in positions of influence using that position to say whatever they think is going to get a rise out of people needs to end. There’s a reason we have social norms and “acceptable behavior” and everything happening in the world is a result of us deciding that’s not important. People should be embarrassed when they offend people. They should have consequences when they say something intentionally harmful. Stop normalizing toxic behavior. Stop rewarding rage bait and grift.

If you’re not using AI you’re dramatically falling behind of what is possible. If you think AI is performing everything perfectly the first time you’re going to drive yourself into a ditch.

@levelsio 4x per capita

Kuala Lumpur has about 4x the green space of Bangkok last time I checked. Makes a big difference. Right now, the air quality is mediocre but it is entirely from burning in Indonesia and Thailand. The religious thing is complicated. Malay muslims have special treatment in certain areas, but it uses common law, foreigners can free hold property, and the vibe changes dramatically from neighborhood to neighborhood. Economically they are getting quite strong as well.

This is true, Kuala Lumpur has never been my favorite place (too religious for my tastes) But I cannot deny they somehow managed to get clean air unlike Thailand or Bali

Similarly, I think @cloudflare is going to be a big winner here as well.

Now seeing both @threepointone and @thdxr talking about how LLMs are better at writing code versus using CLI tools. 🤔

I think Apple might be perfectly positioned for the AI future by simply not spending. Keep making incredible hardware that runs on-device models. Open source is consistently close to frontier models, so why burn billions developing your own?

@sobedominik - Ghostty w/ OpenCode - Neovim

What's your coding stack these days? I feel like with agentic coding it becomes less and less important to be in the IDE for the majority of the time but I'm still defaulting to Cursor because I can use: – Terminal with Claude Code – Git viewer – Quickly edit code I've seen people use things like Cmux more and more and just curious what your stack is these days?

@kenwheeler Nope… right there with you. Big numbers need to go up!

am i alone in feeling like anyone holding openclaw up as some kind of moated innovation has lost the plot entirely

How much longer until asking your boss a question versus running it through AI first becomes the next “just google it.”

@levelsio The dirty secret about lots of AI start ups is that they demo better than they work.

Icon, the AI Admaker, just went bankrupt They paid $12M for the domain Icon.com and now it's dead