If anything, gemini models are the least benchmaxxed out of any lab, IMO.

And the benchmarks agreed with you... until now.

So, yes, maybe it's still not - but this would be the only time it would be highly suspicious / obvious benchmaxxing / obviously bad benchmarks.

Is the Gemini CLI still terrible compared to Claude Code and Codex? The harness the main thing holding back Google models as they could've been the best given all the advantages in compute capacity and training data they initially had, where now even the Google CEO said they're falling behind in agentic tasks, which is sort of a vicious cycle because RLHF relies on human usage.

In May they replaced the Gemini CLI with the Antigravity CLI.

https://developers.googleblog.com/an-important-update-transi...

That was sunset and replaced by Antigravity. FWIW until I abandoned it knowing the sunsetting, I was able to get good behavior out of Gemini CLI with overriding the system prompt. The default prompt crippled the harness with very poor instructions, but there was a hidden ENV to override it. Replacing it with Claude Code like prompts based on the model selected, it ran at a much higher intelligence level full stack with significantly less errors.

There is no Gemini CLI anymore, nor you can use Gemini with your own harness unless you pay per-token.

it's called `agy` now

It was superseded by the antigravity CLI.

it is antigravity now. It is ok

That is not the problem of GPL. I'd rather have high quality software with patches shared back and always being able to see the code than some popular software that eventually is used in some corporate environment without the ability see what they changed.

Regardless, who cares about "adoption?" That's not the original point of open source which was that if you're working on something and you wanted to share it, you could, without any expectation of usage, merely that it might help just one extra person out there. It's simply human ego to care about adoption and GitHub usage and stars especially have accelerated this.

More like, don't expect anything if not forced. Most people only generally pay for things if there is no other option. It's about what's moral (even then debatable as everyone has different morals) versus what's realistic to expect.

The Guardian's funding model [0] proves this wrong.

People will absolutely pay for things voluntarily if you ask them, without forcing them to.

[0] https://www.theguardian.com/membership/2018/nov/12/katharine...

In a nutshell - voluntary funding by readership. No paywall. You can read as many stories as you like for free, and donate if you want. It works.

I've always wondered how they are actually making ends meet, if at all, that is, what their budget looks like.-

That article has no detail. How much did they make? They said they hope to break even in 2019 so they weren't profitable, what are the latest stats? Compare this to NYTimes which has been crushing it after implementing their paywalled subscription.

But when I said the exact same thing [0], people were talking about how it's their "culture" to use corporate laptops for personal usage, and I even got downvoted and flagged if you read the thread (turn showdead on in your HN settings). It's baffling indeed, but I guess people will learn the hard way, that companies can see everything you store and even type on there.

[0] https://news.ycombinator.com/item?id=47851948#47854312

Really depends if company has some spyware installed. My work MacBook was just given me as plain new MacBook that I installed everything myself. Personal iCloud, work OneDrive, no corporate spyware

It’s always good at the beginning. What happens when you leave? Is the laptop yours? Theirs? Is this documented?

I have worked in 3 companies and none of them had installed anything on the work laptop. One IT department tried to, but I ignored all their messages. I would probably be pissed if I have to give up the right to work with privacy.

I had formatted the laptop both while getting it and returning it. I had filevault on and I wouldn't have shared my password.

Modern MDM solutions use Apple Business Manager to automatically deploy the MDM after a system wipe. If you have a company owned Apple device, and the company has it enrolled in ABM, then you can wipe and filevault all you want, but the agent will still be reinstalled, and the file vault key escrowed.

Yeah I know. It's not like Apple is trying to hide it or most companies try to hide spyware, and I know how to check for Apple business manager or spywares.

Off the bat, you said that there shouldn't be expectations of privacy, not just that there currently isn't, which is not the exact same thing.

I've never used company equipment for anything personal beyond a quick Google search, but I'd still like there to be some degree of privacy in case I ever feel pressured to do so, e.g. by a deadline to fill a form.

Mostly individuals decide what they will do, and then communities decide if they respect that or not. Doesn't exactly apply to companies tho, since they're functionally sociopaths.

As said elsewhere:

Chain of thought does not exist in the output of Claude, they disabled true thinking due to distillation risk. What you see when thinking summaries are enabled are just that, summaries of thinking into Claude-isms, therefore you cannot make any inferences on what the model is doing unless you literally work at Anthropic and can see the true thinking traces.

Of course you can make inferences what the model is doing. The summaries are usually sufficient. They're summaries, not random noise.

Sure, but that doesn't tell you about how the CoT is phrased when the agent is its own target audience, which is the interesting thing under discussion here.

I remember enjoying watching Fable think during the original limited preview. It was full CoT for sure. They must have removed that feature recently.

I use open models for non work stuff and sometimes I cancel the output because the CoT is all I needed to read.

Same. (And/or interrupt the process and save time as you see it diverge by getting your answer wrong or going on a tangent ...)

Chain of thought does not exist in the output of Claude, they disabled true thinking due to distillation risk. What you see when thinking summaries are enabled are just that, summaries of thinking into Claude-isms, therefore you cannot make any inferences on what the model is doing unless you literally work at Anthropic and can see the true thinking traces.

Yes. I'm talking about what's in CoT generally, based on various rumours, experiments people did with previous models, stuff in the recent METR report on the HF hack, etc.

Increase public transit density. In Asia like in Seoul or Tokyo I'm not further than a 10 minute walk anywhere to a public transit station.

Improve public transit then. This is not a good argument to instead allow everyone to have cars.