Hacker Newsnew | past | comments | ask | show | jobs | submit | Dibes's commentslogin

I just recently started using T3 Code and it's strongest points are having a really good workflow, interconnecting all agent harnesses so you can easily orchestrate across them and they can easily communicate, and truly phenomenal remote sessions. All while having a singular UI/UX experience which I actually really appreciate.

It is probably weakest at customizability and making sure you have parity in all your agent harnesses with skills/etc, granted I just set an agent to ensure that so the point is a bit moot.

Overall I have really been enjoying the inbox style workflow and it has been a joy to work with. It jives with how I like to work across streams.


To restate what I put elsewhere, I genuinely don't think I care much for how it was written. The problem is that it felt exhausting to read and was chock full of the same claudese voice that makes claude so infuriatingly exhausting to read. I gave it a try and gave up halfway through. If it was AI generated but was a good read that conveyed a central point/theme, I wouldn't be complaining. This at least is trying to argue a single point, but it does so in what feels like at least 10x the word count.


I usually don't complain about stuff being AI generated. But in this case, regardless of HOW it was written/generated, it reads exactly like claudese and frankly I am so exhausted of reading that particular voice that I made it half way through this article before I just couldn't do it anymore. I really don't care if AI was used, I care that it is readable. And frankly, it isn't.


"That particular voice".

Yeah. Different authors should have different voices. Having them all sound the same is like having all your food taste the same. It makes you want to throw up after a while.


I think you missed the point, "That particular voice" - is the claude vomit that I have to consume everyday and it makes me shudder every time I see it.


No, I don't think I did.

That voice grates on you. It grates on all of us. But it would grate on us less if we didn't hear it all the time - if it wasn't half of the words coming at us.

Dibes said, "I am so exhausted of reading that particular voice that I made it half way through this article before I just couldn't do it anymore." I think what I said is exactly to his point.


I'm not sure what to make of this graph[0]. It shows medium as the most effective thinking mode by far for frontier code.

It's the only case that I saw going through the system card where more reasoning effort meaningfully negatively impacted the resulting eval. I know sometimes max efforts show a small dip, but this is substantial. I wonder why in the world that is?

[0] https://imgur.com/a/Nv8V7Ry


I'm not sure about the answer here, but this can be caused by the scoring rubric used by given benchmarks. For instance, if a benchmark docks scores for running too many commands or using too much wall-clock time, higher efforts will get lower scores.


Pure speculation, but I've noticed drawbacks to the models on high effort. I interact mostly through prompts rather than agents so I sometimes see where their reasoning falls short. A model on high effort has longer output and can get hyperfocused on irrelevant details, maybe increasing the surface area for mistakes. I haven't used other effort levels extensively yet but I've supposed that medium may have more balance.


Very interesting. I mostly interact through prompts too but haven't changed the effort that much. I just assumed higher was better at the trade off of tokens so I have largely left it at the default.


It's "Cost per task", so perhaps it burns tokens too quickly, trying to "do a better job". Over-engineering :)

Last week it felt like Opus 4.8 was moving the Pro "usage" meter very quickly. Today, pre-announcement, Opus 4.8 Medium felt like there was less meter-use per minute. And post-announcement, Opus 5 Medium also feels more efficient, allowing more work in the 5-hour window.

Completely subjective, of course.


Just read something, my subjective view was correct, but it's not the model:

"Your weekly Claude Code limit is 50% higher through August 19"


apparently it got docked points for editing files out of scope


This must not be weighted very heavily on the benchmark because if it was, Opus would bomb every test (half kidding)


Do you have a source for this? That would explain it, but could be a bit of a concern on the general focus the model at higher thinking exhibits.


Maybe it's akin to the Ballmer peak: improved performance at a specific level of relaxation


I agree, very odd they did not comment on any theories for the degradation here. Dip and then rebound at max effort is pretty interesting too. Overthinking is bad, but you can overthink so much it starts to be better again?


It is probably from randomness. The benchmark tasks nowadays are so long that you can't really afford to run a large number of samples of them per model & effort combination


No it's a mean of 5 runs.

> We report FrontierCode’s overall score, a composite measure that grades each patch on blocking functional criteria (held-out unit tests) together with weighted code-quality rubric criteria, as mean@5.

They don't explain more in the system card, I guess higher effort levels could loose points on the code quality / scope / style / maintainability stuff?


If you're doing passes@1, especially for long-horizon agentic benchmarks, you might as well as not do the benchmark at all.


Odd that open AI models aren't on that graph but are on the rankings! Must be a data lag issue?


For me claude is the best, hands down. Fable took it a step even further.


I agree. Gemini actually is pretty good for isolated components too. But fable is much better at design than opus or gpt5.5. I have not seen as much difference elsewhere, but definitely design fable is great.



After clicking play, there's a prompt to log in. I had to use a video downloader service to watch it


You can just replace x.com with nitter.net, though their bandwidth sucks for media playback.


Harry Potter would have posted the original video to YouTube instead. It seems like a tragic irony, and perhaps a sign of danger, that OP posted the demo video to the website of He Who Should Not Be Named. Why? Why?!


Because Twitter posts consisting of YouTube links die quickly. It’s very obvious once you have a few thousand followers.

Your option is basically either upload to twitter, or put the YouTube link at the end just before a screenshot. Or both a video and a YouTube link, I suppose.

If you trigger their YouTube embed, it seems like it gets penalized quite harshly. I’ve seen other people agree with the sentiment.


Just don't use a Twitter post as your demo video in the GitHub readme. I don't care what someone does on Twitter - I never go there. But you can just embed a .gif or .webm in your readme or link to YouTube there.


Twitter posts never die if you never post to Twitter in the first place.


What does it mean for a post to die, and why is that bad?


The poster to x apparently has a bsky account also — I wish they would cross post.


or xcancel.com


that was my first move, but i got an infinite redirect


could be regional cuz that don't happen in the US. they play right away regardless of log in status


I'm in the US and twitter videos don't play for me. I see what looks like a video control when I view the post, when I click it, it throws up a modal log in/sign in dialog. No way to view the video while logged out.


I don't have a twitter or x account. I clicked the link, a log in modal popped up, I dismissed it, then was able to play the video. Firefox on iPhone if it matters.


> Firefox on iPhone if it matters.

very possibly, I am using Chrome on a desktop.


That's simply not true.


They used to. Not anymore.


I believe they were trying to push back on the easily word in "easily produced". While, thank god, modern medicine makes birthing dramatically safer than at any time in history, it is still an incredibly taxing process on the human body and still fundamentally dangerous. It is not "easy" per-se.


C’mon now, let’s not be intentionally obtuse to try and sound smart adding “yes, but” to a perfectly solid argument. Children are easily produced, that’s a fact. The process of child bearing is not in question, but rather the availability of willing partners (hell, even unwilling, if we’re getting that dark) and how little effort it generally takes to get the conception process going.

They were making a point that it’s much easier to generate a human being than it is actually generating economically meaningful value. One needs a few minutes of your time and gets you feeling really good at the end of the process, and the other can (and often does) require years of effort, toil and sweat (all of which can catastrophically fail rendering your efforts moot).

Generating a new human is, indeed, comparatively easy.


> One needs a few minutes of your time and gets you feeling really good at the end of the process

Ah yes, and after the intercourse the child spawns, I remember. And then it just leaves and goes somewhere and starts it's own life. No pain, work, food, time or whatever involved. Just having sex and feeling good.

> years of effort, toil and sweat

Would be a shame if that would be needed to produce a human being.


Those points are fine, but not the root of what makes MAGA shameful. You can go about having that opinion and take actions towards it without being racist, anti-LGBT, generally hateful, and backing an administration that has been proven time and time again to be deceitful in every facet and tuned to the interest of the wealthiest.


The thing is, for most people a standard wall outlet is plenty. The math works out that a simple 15/20 amp circuit can charge over 40 miles overnight, and the vast majority of people aren't driving more than that for their daily commute plus errands. Level 1 charging is genuinely sufficient most of the time. I was particularly swayed by the technology connections video on this topic I watched before buying my first EV https://youtu.be/Iyp_X3mwE1w

The real pain point, in my opinion, is whether you have any place to plug in nightly. If you don't, then as you pointed out, it becomes a nightmare to own. Range anxiety is completely justified when public charging infrastructure is still as unreliable as it is, years after the initial build outs. Your points about charging pain are all too common.

If you have a garage with an outlet, you are generally fine. I lived off a level 1 charger for over a year before I decided I wanted the convenience of a level 2 charger.


Sorry, I love Technology Connections as much as anyone, but that's a ridiculous argument. Even people who drive less than 40 miles a day will occasionally need to drive 100 miles a day for two days back to back. That's not even a long distance trip, it's just driving around. With level 1 charging they are stuck and frustrated. With level 2 they're fine. Not to mention the hassle and mental energy required to plug in and out for every little trip.

For most people a 240V outlet is worth it. Not to mention it's at least 10% more efficient, which is quite significant and weird that Technology Connections didn't mention that.


You can use DC fast chargers to fill gaps as needed.

Also if you are always driving 40mi/day, you likely float with a battery percentage around 80%, leaving plenty of capacity for those consecutive 100 mile days with your standard overnight slow charge.

Again, this cannot be said enough, EVs are not gas vehicles, they do not refill like gas vehicles, if you apply gas vehicle logic to them, they look awful. But they are not gas vehicles, they don't follow the same logic and rules of gas vehicles. So you don't apply gas vehicle logic to them.

It's like handing chopsticks to an 18th century westerner, they'll stab their food with it and laugh about how stupid and useless they are. You need to learn and use chopsticks before criticizing chopsticks.

This whole thread (as always) is full of people stabbing their food with chopsticks.


Look, I don't care, I know there are strong opinions about how these discussions sway people one way or another. I'm as much of an EV technology fan as anyone, but I'm speaking from personal experience with this exact situation: if I didn't have a 240V charger in my garage, my EV experience would be garbage and I'd give up on it in frustration. I own one of the most common EVs, I have DC fast chargers in my area, I don't drive my EV that much during the week, but when I need to drive a bunch of short trips on the weekend, this exact scenario arises. I don't care what your theoretical model of an average EV driver looks like, I'm telling you that it doesn't match my reality and I am certain the reality of many others.

What's bizarre is that this should be incredibly non-contentious when it comes to EV adoption. By code, everyone in the US already has two phases at their panel and running a wire and outlet in their garage (or a weatherized cable to the outside) costs $100-150 in materials and a similar amount in labor. This is literally negligible in the broader scheme of the automotive economy. My humble suggestion to you is: save your breath, we're on the same side, raise your voice instead when it comes to demanding a sane EV industrial policy, regulatory policy, urban planning policy, removing subsidies for oil and gas industries, and the like.


That example doesn't make sense, because 100 miles back to back is only 200 miles. You've got about 80 miles from charging overnight those two days and another 200+ miles already in the battery. In that situation you're totally fine. After that there are superchargers of course.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: