Maybe he's not high level enough employee to have any say in the product roadmap, and he's behind on leadership planning?

I don't think that's quite it. They had it on their website before this, that opus 3.5 was coming soon, now they've removed that from the webpage.

Also, Gemini ultra 1.0, was released like 8 months ago, 1.5 pro released soon after, with this wording "The first Gemini 1.5 model we’re releasing for early testing is Gemini 1.5 Pro"

Still no ultra 1.5, despite many mid and small sized models being released in that time frame. This isn't just an issue of "the training time takes longer", or a "skew" to release dates. There's a better theory to explain why all SoTA LLM companies have not released a heavy model in many months.

Are you also implying that new 3.5 sonnet is better at those things?

No, Opus is better. I have no experience with 3.5.new.

I think the main reason is they tried training a heavy weight model that was supposed to be opus 3.5, but it didn't yield large enough improvements to 3.5 sonnet to justify them releasing it. (They had it on their page for a while that opus was coming soon, and now they've scrapped that.)

This theory is consistent with the other two top players, Open AI and Google, they both were expected to release a heavy model, but instead have just released multiple medium and small tier models. It's been so long since google released gemini ultimate 1.0 (the naming clearly implying that they were planning on upgrading it to 1.5 like they did with Pro)

Not seeing anyone release a heavyweight model, but at the same time releasing many small and medium sized models makes me think that improving models will be much more complicated than scaling it with more compute, and that there likely are diminishing returns with that regard.

Until you have reform in the sense of replacing the First-past-the-post voting with a more representative system that allows the survival of more than 2 centrist parties, or have a system where there are meaningful referendums at a meaningful frequency, then where is the value of being "politically informed" on new/current politics?

Democracy in many western nations, at least in the US, is more or less an illusion of choice. Being sucked into the liberal/republican squabbling, drama and even the occasional political issue is nothing more than mere entertainment for the peasant class. For lobbies and corporations, who actually have much more leverage into governance, then yea being politically informed for them is prudent.

"we process the tweets directly in your viewport as you would have seen them."

The primary issue with this approach is, even with the speed of gpt-4o mini, often times you're going to be displaying the "harmful" content for enough time for the brain to process it. This is especially true when you're dealing with images and short 1 sentence content like twitter. I think you'll want a safety mode, where nothing is displayed/or you have a css-blur on it, until it has been vetted.

Forgot to mention we actually process ahead of the viewport as well so other than the first load it's highly unlikely you will see an unfiltered tweet (from personal experience of using it). But thanks for the idea at the end I think I'll add that!

technically, you can add the tag "woke" and hope that Gpt-4o mini's definition of woke is the same as yours. Doubtful it will give you the results you're after, but maybe to a small extent.

The free plan is extremely limited, just runs on one website and only blocks a 500 tweets/month. Also gpt-4o mini isn't that expensive to be $5/day/user, even with a lot of page views.

Yeah I think I'll make the free plan better, thanks for mentioning that.

I was interested in giving it a shot, but i didn't see any setup instructions or anything i can download from that page. Is your extension available for use?

It isn't, the blog was just a high level recap of what I built, as I unfortunately don't have enough time to turn it into a product that can be used.

I'm very happy to see that someone is building it though, I really think that personal AIs/LLMs are the future.

You should still slap the code on github with a disclaimer that it's not stable/fully featured.

I like the concept, but i wasn't able to get it to filter anything, even when i used a tag, "technology" that my homepage is filled with and it didn't filter anything (I did a "Save options" and a page refresh, tried the power button in both states (it's not obvious to me, or chatgpt-vision which is the on-state, but regardless i tried both and nothing was blocked)). Any troubleshooting tips? (and set a todo to fix the on/off button ui)

unrelated: for youtube are you classifying the content purely based on the title? or are you digging deeper and looking at the description and thumbnail and comments? Is there a second where the content is visible before it is blocked/hidden?

Just checked your account, if you could press the button on the top right and make it look like the power (without the slash) and then save? since it looks like it isn't enabled on your end.

And yeah still very early will be improving the UI a ton and taking into account your suggestions for the next version.

In terms of youtube, currently we just take into account the title, sometimes it is visible before hidden/blocked but we usually process within ~700ms so vast majority of the time you wouldn't see it.

I tried again, and now it's working - and yes it is reasonably fast but i wonder once you start adding vision it will likely increase your response time, to the point that you may need to consider a temporary blur until the content is vetted.

Anyways good work on the tool, it has potential even at just blocking inline ads and spam.