Hacker Newsnew | past | comments | ask | show | jobs | submit | embedding-shape's commentslogin

To be fair, the post does start with "This is a guest post by Timothy Gowers", feels pretty clear already. But of course, assumes reading beyond the title.

For reference, for the ones who aren't aware of the constant drama that seems to follow V: https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que... (open up almost any discussion and see the flood of similar messages).

I never paid enough attention to figure out if the drama is justified or not (and don't really care either, please don't try to convince me to care either, don't shoot the messenger).


There’s good kinds of drama (competent dev forcefully rejects bad ideas) and bad kinds of drama (activist/corporate takeovers, subpar dev doubles down on mistakes, etc), but I haven’t been following it, so I have no idea which one this is.

I feel like they're being outright misleading unless they publish the actual transcripts.

We have zero idea what the prompt was, what OpenAI provided, how the model arrived there, and sharing that quote like "Look what the model came up with!!1" without explaining the background and context, feels like it's intentional so they can claim "These models really are acting by themselves" rather than taking responsibility for their fuck ups when it comes to the security testing.


AI optimists getting hunted for sport in 2085:

"lol this is either a marketing ploy or just negligent security testing"


I was gonna ask how people found their coding plans, and realized, have they massively ramped up the prices? Seems the middle plan is ~$80/month now, didn't that used to be like $20/month? Cheapest plan is ~$20/month currently.

They must have hit really hard scaling limits if the prices were hiked so much so quickly.


I paid $360 annual for Max plan and currently averaging about 1BN tokens a day with their frontier GLM-5.3 model. This was clearly unsustainable for them and they've dropped this package.

1 billion tokens a day?!! I've done a lot of work these past 2 weeks with GLM-5.3. Like, a lot. And I've just passed 300 million tokens in total.

Can I ask where are you using all those tokens?


Well, there's essentially two major ways to use these models: Pair programming or fully autonomous fire-and-forget code generation. The second strategy needs essentially zero input, so the number of tokens you can blow is practically only limited by API speed.

There's also a third way that can spend the most tokens: if the AI is used as part of the product, and not just a tool to build the product.

Something like this I guess: https://youtu.be/U-Rqv9dOB1U

This was such a gem of a video.

This is such a good video. Instant sub. Next to tech bros, we should also put AI-cringe bros.

I have 3-5 agent harnesses with large context windows working on different applications concurrently.

Share the resulting code from any one of those please? I've tried so many times to find a setup that facilitates parallel work + high quality results, but it's just impossible regardless of harness or model. Leave the agents alone for too long, and the entire thing just balloons out of control, and next you know you're sitting there with half a million LOC where 80% isn't even needed.

Most of them are not public, but a fun thing I did was a mario cli game - https://github.com/Daviey/mario/ (or `ssh mario.baby`).

I now exclusively use https://omp.sh/ as my harness:

I set it up so it never works in the main branch so subagents etc don't step on each others toes, and only merges back when complete: https://github.com/Daviey/mario/blob/main/.omp/hooks/pre/wor...

A good AGENTS.md is essential: https://github.com/Daviey/mario/blob/main/AGENTS.md

I then provide specifications for what I want, making sure it is unit tested.


That's easy to do with many agents independently told to find bugs in a large codebase.

300M for two weeks is surprisingly low. What are you doing that need so few tokens?

It's not my main model (that would be Fable 5.1 Extra) but it's been doing agent-driven search and optimisation of a cross-trading ranking model (it's for work).

I would suggest you to hook fable or 5.6 to check it regularly and its work because it gets lost easily on stuff it was not trained on. I'm doing some custom inference engine optimization and it's a workhorse but it can easily lose its way and if you don't recheck it you will get wrong answers in the end.

Yeah that's what I already do. Fable writes the plan and checks things at certain milestones. Otherwise it does get lost indeed.

Kind of feels like this applies to every single model, from Astra to Qwen, they all eventually lose track of the plot unless you feed it some human's input that can steer them right every now and then. The only difference is how often you need to do so, and also how often you want to do so heavily influences how good quality the results will be.

you are not false, but there is still difference. its just that the better models are correct more of the time and will better validate its own steps. glm sometimes will understand the plan start implementing and then forget part of it and then say it finished. or then take a wrong turn somewhere and not correct. but they will all happily proclaim they are correct till you question it.

I also have a legacy pro plan and the only limitation is if you are trying to work in the morning from Europe because you are in the 3x usage overlapping China time but after 12 or so you basically can run it at least for me at least 3 parallel sessions all the time.

>I was gonna ask how people found their coding plans

Very good - but I'm on a legacy plan. And coming up on a renewal that would put me on the watered down current plan. But with 50% legacy discount think it may be worthwhile. If I go to a competitor I'd be paying market rate.

>They must have hit really hard scaling limits if the prices were hiked so much so quickly.

Not really scaling - their plans were initially comically subsidized even more so than what the western providers are doing. More advert for an upstart than commercially priced.


Their plans are still worth it if you use their models. You can see how many tokens you can except to get based on plan here: https://docs.z.ai/devpack/overview#estimated-token-allowance

The max plan will provide ~1,100 USD of GLM-5.3 or ~260 USD of GLM-5.3-flash per month for 168 USD. I can personally attest to these numbers through omp (~97% cache hit rate).

Unless you are able to highly parallelize (your work, you won't be able to hit your hourly or weekly quota using the flash model simply because it's so slow.

They give you ~3x more flash tokens, which maybe comes out to ~2x more actual work after accounting for the extra thinking it does to achieve the same result. The mental model, for not getting angry, is 5.3 is fast mode by default, and you can disable fast mode for 2x the work output at 1/3-1/10th the speed.

They're serving me 5.3 at ~40 tok/s and 5.3-flash at 30 tok/s (according to omp).


That table assumes cache hit rate of 95% or better. Am I understanding this correctly that people really are doing such repetitive prompts (compared to each other, across the concurrent user base at that time) that only 5% or less need actually be computed by the intended LLM?

That is shocking. Is it per-token I wonder?


Every tool call is essentially entire prompt so far sent again with the response and that's why cache rates are so high for agentic workloads. This really bites when using expensive models since most models are 1/10 for cached input.

We have been running a lot of agentic benchmarks with the various loops and tool calls on longer threads - we routinely see 90%+

Just checking now: recent runs tau3[1] was at 96% and toolathlon[2] was at 90%

[1] https://www.induction.ai/docs/benchmarks/tau3 [2] https://www.induction.ai/docs/benchmarks/toolathlon


If you are using their coding plan for coding, then yes you can easily hit such cache rates, with a good harness.

I’m getting 97%.


The way I look at it, their coding plan doesn’t retain data or use it for training making it one of the cheaper plans for me.

https://docs.z.ai/legal-agreement/privacy-policy


You believe any of these companies care about the law? They care about winning and building the self improving AI as quickly as possible.

I too am sceptical but I’ll take my chances. At least it’s helping the open weights.

I believe the that the companies who claim to not train on my data are more likely to not train on my data than the companies who refuse to even claim they won't.

Also why Meta gets a +1, just charge less money on the training path.


I’m not sure that follows. You’re assuming that all those claims have the same weight, without considering the size, jurisdiction, reputation or even the general vibe of the company making that claim.

If you factor that in, then there are clearly different tiers: one you can trust, and one that may well just be saying that to increase market share with little reputational or legal consequences if they are found to be lying.

These are not equal.


Yes I sometimes think the "don't train on my data" is actually a good signal for "this data/person is probably better to train on because they want to keep something private". The whole copyright system should have stopped these guys from training on everyone's data and it did not, if you think they care about the privacy checkbox I think you're dreaming personally, based on their past behavior.

> I’m not sure that follows

To be fair, none of us are sure of anything and I think that’s the part that’s most irritating


It’s more a polite way of saying “that’s crap”

And mine a polite way to say “you are equally uninformed”. We’re not getting anywhere. All the best.

FYI it’s helpful to actually say your point during a discussion. And if you don’t want a discussion then why did you comment?

Way to restrictive in terms of tokens provided. I am on their largest plan, and quickly run into their limits. And that is using it selectively in addition to codex.

Yeah it went from a great deal to unviable compared to other providers imo. They really need to find a healthy middle ground

It just gives a taste of what we are all going to have to pay soon, once the model providers actually have to make money. And the era of "let's charge a dollar for every 10 dollars running the infra actually costs" is rapidly coming to an end.

And you can bet GLM is still ridiculously subsidized, just not as ridiculously as Anthropic and OpenAI.


This isn't true, you can pay for GLM 5.3 from a provider like Neuralwatt or Friendli who have no incentive to subsidize or loss-lead their inference APIs

They didn't pay for training

This introduces other incentives to cut corners and over-quantize.

What provider are you using currently?

It's hard to know, since no one advertises the actual token limits (partially cause they're prolly complex / adaptive). So it seems much more likely that they just offer different pricing tiers than you're used to. Like, the $80 plan is still ~$80 of subscription quota, regardless of what else is offered.

For [API usage](https://openrouter.ai/z-ai/glm-5.3-flash#providers) they charge a bit more than the very cheapest providers of GLM-5.3-Flash, but not so much that a big price difference would make sense.


ctrl+u generally works for lots of password prompts to reset/remove what was typed, wherever you come across a password prompt and a keyboard including those two keys.

> The best thing for long term browser compatibility is focusing resources into Blink.

Lol, for Blink perhaps, what about the rest of the ecosystem, the web and the users, like the ones who don't even use Blink?

Very idealistic point-of-view with no basis in reality, sadly, because wouldn't it be nice if we could care about just one engine?


Was this recent? Because Firefox were the browser that worked better than all the rest, for quite many years, maybe 2005-2010, and Firefox had no competition basically until Chrome arrived on the scene.

Just the fact that Firefox got browser tabs before anyone else, and Firebug, just took it to another level compared to all the other browsers available at the time.


Your history is off. Opera was the first mainstream browser to implement tabs and boy was Opera also much faster than Firefox at its peak (both UI, resource usage, and rendering) which overlaps that “peak” Firefox period by a lot.

Firefox was great because it was open source but it wasn’t ever really the “best” browser - it was the best open source browser.


> it was the best open source browser.

Yes that was implied because IE was not mentioned either!


When the original Phoenix browser was released, Explorer was in the middle of its slumber with the dreadful v6. Phoenix might not have been as feature-full as IE6, but it was very much the superior experience. It made browsers fun again.

That’s a quote you seem to have materialized from thin air - op didn’t make this claim.

It’s literally the last sentence?

not sure what you’re quoting but the last sentence of the post I replied to is:

> Just the fact that Firefox got browser tabs before anyone else, and Firebug, just took it to another level compared to all the other browsers available at the time.

Your quote doesn’t appear.


> Hardware accelerated video decoding (for video playback or for WebRTC) is available via VA-API for both X11 and Wayland. Since firefox-101.0.1-4 package it's enabled by default for Intel/AMD users.

From https://fedoraproject.org/wiki/Firefox_Hardware_acceleration

I think that was around 2022-23.

I remember this affecting my laptop video watching a lot. Google meet feeling sluggish on Firefox is also something I remember until very recently..


Google still tries to sabotage Firefox experience on Google Apps, especially Meet but also YouTube and sometimes Google Sheets/Docs as well.

They want to degrade the experience just enough to get you fed up with FF and go (back) to Chrome, without making it absolutely obvious that's what they're doing.

They often label these things as "regressions" because they don't test on Firefox and take forever to fix them. But I can't help but believe it's intentional.


Yes, exactly. Google products working worse on Firefox is a problem with Google, not Firefox. I'm not sure why the conflict of interest isn't more obvious.

Firefox was bad on linux for a long time. Even running the windows ff build via wine/proton got you better performance than the "native" build (I think that was around 10 years ago).

Much longer than 10 years! I’ve been on Linux since 2008 an it was never a problem.

IE5 and then IE6 were head-and-shoulders better than Mozilla around the 1999-2002 era. They were significantly faster and lighter, and not just because of the deep OS integration but because Mozilla's browser was actually a full Web suite, with a built-in page editor and a mail client.

Then Microsoft stopped developing IE, and it became essentially frozen.

Mozilla moved ahead with refactoring the browser parts into a lightweight Phoenix (Firebird, Firefox - anyone remembers the FireSomething extension?). But it took _years_ for it to get into a usable state.


So right now Servo is funded by the community, and this is no good because why? Instead you propose that this project should be funded by a for-profit company with it's own objectives in this world (that often starts with "more money"), again leading me to ask why?

The setup right now seems to give enough people enough time to work on it to move forward, why on earth would you want to involve companies infamous for their shit software quality, like Samsung?


> So right now Servo is funded by the community, and this is no good because why?

Because look at Mozilla... You need to get the browser in front of people. A browser that works on the real web, is such a huge project, you cannot do it as a hobby community-funded project for ever. You need user-share-%.

Samsung can put it in front of Samsung appliance users.


Samsung has put Enlightenment 17 in front of every Tizen user. I am not seeing the flood of apps implemented on EFL coming out

I think it is actually Mozilla people look at when they ask "why" :-P

Probably because broad adoption would help making it a real option to use in the near future?

So sure, Samsung would bundle it with their usual crap, but if it helps making the engine more mature, great, but I also don't see why they would in the first place.


It had broad adoption, Mozilla threw it away.

The same would occur again.


In fairness, while there was definitely Mozilla mismanagement, there was also a massive monopolistic marketing/bundling campaign by Chrome. Google search and other properties pushed Chrome, Google paid application publishers to bundle Chrome the way they bundled adware, and so on.

Servo never had. But yes, mozilla management let FF go down the drain.

Down the drain? Firefox is still a great browser.

Their performance is not great comoared to the competition and their market share is at 3.7% and falling. The only thing that keeps me there, is ublock origin. The moment ladybird or a a servo base browser becomes an option I switch.

And if FF also starts to restrict UO, I just go to chrome then. I assume many think like me.


Firefox is acceptable, but it's definitely not what it used to be. I've long stopped expecting any positive user- visible change in Firefox updates.

Servo was not intended to replace Gecko. By all reports it was an experimental component, many of whose findings were used in Gecko.

That folks want to take it further is admirable. But doubtful Mozilla could have afforded to with their existing commitments.


sorry, what? servo existing as a mozilla side project did not give it broad adoption

> Instead you propose that this project should be funded by a for-profit company with it's own objectives in this world (that often starts with "more money"), again leading me to ask why?

You understand that Mozilla Firefox is not funded by the "community" and is completely funded by Google's money. Even donating does not fund the browser.

Mozilla knows that the open source "community" has no money, and cannot replace the amount of money Google is paying them and the employees who work on the browser.

The question is are you ready to fill the $600M a year black hole for Google to NOT pay Mozilla to fund Firefox?

It only makes sense for Servo to get funding from for-profit companies at this point, which is one of the only ways open source can work.


I prompt a lot, but I also still run a lot of commands and do quick edits with vim myself, not gonna prompt "increase X to Y" when it takes ~3 seconds to do myself vs ~30 seconds for even a fast model to do it.

https://news.ycombinator.com/jobs

It's a talent acquisition ploy.


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

Search: