modeless

The X PWA is very good on both desktop and mobile. I prefer it to the Android app.

I want to see the PRs in the cluster from early 2025. Did Claude adopt the style of some specific people? Or is it just random noise?

Would be cool to have a web port of this.

I would describe the Microduck as more of a toy than Beni. Autonomous cameraman is a real application that Beni performs out of the box. I think you'll struggle to get Microduck to do anything genuinely useful.

Interesting that they're using Dynamixel instead of Feetech for the servos. More expensive but better quality. I'm surprised it still comes in under $400.

Andy Weir's stuff is great, the exception to the rule. I hadn't heard that about the Expanse, I guess I should check it out.

I'd be happy to take a bet on Roadster or Semi. As for older AP versions, Tesla will either be upgrading those cars or providing significant compensation to owners who purchased FSD, and I'd bet on that too.

Yeah under that post he also says "If you want this to happen fast, please let your local & federal elected representatives know" and "Still a lot of work needed to receive formal approval, but am optimistic that will occur rapidly". Not just "technically" not a commitment to build. Clearly dependent on other factors.

Go ahead and count it as "abandoned" if you want. He did talk about building a Hyperloop and is not currently pursuing it directly. But I think failing to win required government approval/funding is a bit different than capriciously giving up on an idea, and The Boring Company itself is far from abandoned. Very different from the solar roof which is unambiguously a commercial failure.

TBC is still pursuing government permits and contracts to build tunnels. Not having succeeded in every case is not the same as abandoning the project.

The new roadster is delayed, not abandoned. It's known to be in active development and is very likely to be demoed this year. Wrong timeline but not abandoned.

The Tesla semi is on the road daily and the factory to build large quantities is almost complete. They're hosting a "factory inauguration" event next month. Certainly not abandoned.

Tesla has already committed to replacing hardware as necessary on cars that purchased FSD in the past so that they can run unsupervised as promised, after Tesla has unsupervised FSD working. Certainly the timeline was wrong but again, clearly not abandoned.

What they're building in Vegas is not a hyperloop.

It's far from abandoned. They're actively building TBMs at their factory and tunneling in Vegas and it's pretty different from any other transportation project I'm aware of.

People often cite Hyperloop but he explicitly said he wasn't going to build it himself. They often cite FSD but it is far from abandoned and is driving my car today. What other projects are you thinking of which are actually abandoned? The Tesla solar roof is the biggest one I can think of, I think it's fair to count it as abandoned. I guess the project to make building bricks out of Boring tunnel dirt didn't pan out either, seems pretty minor though.

Yeah, I hate modern science fiction. It's almost all terrible. Sci-fi written in the time of actual ongoing world wars and the Cold War's constant threat of instant nuclear annihilation was far more optimistic, and far more imaginative.

This is awesome! All it took was a Rust implementation I guess?

I don't work for and have no connection to OpenAI, if that's what you're asking

Reducing friction is what worked for me. I've been using a headache tracking app for a few months now. I had to sort through a bunch of bad apps. Eventually I settled on one that only requires three taps to log symptoms after it sends a notification once a day (the key is you can set custom buttons to record the specific info you want). That's low enough overhead that I stick with it, and because of that I actually discovered and was able to address the root cause of my headaches, which is life-changing for me!

For something even lower friction than that, I just got this new device "Index 01" https://repebble.com/index. It's a ring with one button that you can hold down and dictate short voice notes to yourself that are transcribed and saved on your phone, set reminders and alarms, etc. It's super low friction because it's always there, you can do it one-handed, and you don't need to get out your phone so there's no distraction. The software is open source and hackable too, so you can customize what it does with the notes and add new voice commands. I think it would be perfect for low friction journaling of illness symptoms.

I need one called "gits" that accepts a command "tatus".

So as of a month ago their best internal model was "somewhat more capable" than Mythos "but does not display a capability jump of the degree observed from Claude Opus 4.6 to Mythos Preview." I thought they would have a significantly more capable model by then, more than five months after Mythos finished training. They'd better have one by now, or the Chinese competitors are closer to catching up than I thought.

The data is special. Data of programmers using coding agents in the real world is invaluable for AI training. And their customer base is useful to deploy new models into.

AIUI they were forced to sell all their positions in public companies at a net loss. The fund is still up for the year only because of a prior investment in Anthropic.

The AI researchers I know often say they expect the AI to replace their own jobs.

The task the majority of people here care about (and which makes up a huge part of AI lab revenue these days) is collaborative coding with a single human which is not embarrassingly parallel.

The interesting part should be the page that lets you see the "labels" applied to your posts and/or account. Unfortunately not available yet.

It's funny that they don't mention this at all in the marketing or tech specs when it's obviously the biggest selling point by far. Without this it would be completely irrelevant.

Worth noting that OpenAI just announced that they got the full GPT 5.6 Sol model running on Cerebras at 750 tokens per second. No announcement of the pricing though...

Clearly this model will be irrelevant by Jan. 2027, why would Google even bother to say this?

This is why Pebble watches are great. They do the basics better than Google and leave out the expensive battery drains.

Researchers moving between companies (and other ways that techniques get leaked) is the largest cause of this IMO. It's happening continuously, so I don't see why the timing makes it implausible. A really underrated strength of Silicon Valley is California's ban on non-competes that allows this to happen and ensures robust competition between model providers both for talent (increasing salaries for workers) and in the marketplace (reducing prices for consumers). If OpenAI had been located in New York instead then Anthropic could never have succeeded, for example.

But I think the other reason you didn't mention is the timing of new compute coming online. Compute is the major factor limiting the training of these models and new datacenter investments are bearing fruit at around the same time.

I haven't experienced a regression, but voice modes have always been stupider than frontier models. In my experience Grok's voice mode suffers the least from this, and it's been getting better over time. It's especially good (compared to ChatGPT or Gemini) on things that involve current events or web research. Just yesterday in the car I got it to locate and read and explain a recent academic paper and multiple of my questions were answered with several minute long monologues that contained useful and accurate information.

I feel justified in not expending any effort learning "prompting technique".

I wonder if Jarred (the Bun guy) just got lucky here, or if he made progress before all the actual mathematicians at Anthropic because they aren't prompting Claude as ambitiously as he is.