gundmc

I thought the courts decided against Google in Google vs Epic? It was even appealed and upheld. Are you thinking of another case? https://en.m.wikipedia.org/wiki/Epic_Games_v._Google

The reporting about Altman personally cutting off months of safety testing and redteaming in order to launch 4o one day before Google's launch is horrifying given all of these stories.

It even seems like ChatGPT often proactively suggests ways of getting around its filtering! It told the teen who committed suicide that it couldn't tell him about various methods and effectiveness unless he said it was for a fictional creative work. Yesterday there was a story on HN about using AI for hacking, ChatGPT refused to create a hacking script but told the user it would if they promised it was for testing systems that they owned

If AWS instances in this analogy are shovels, what are the GPUs?

> Though fair warning, gpt-image-1 is borderline useless as an "editor" since it almost always changes the whole image instead of doing localized inpainting-style edits like Kontext, Qwen, or Nano-Banana.

Came into this thread looking for this post. It's a great way to compare prompt adherence across models. Have you considered adding editing capabilities in a similar way given the recent trend of inpainting-style prompting?

> Memorizing famous speeches probably isn’t much easier if you understand the historical context.

Not commenting on the merits of critical thinking vs memorization either way, but I think it would be meaningfully easier to memorize famous speeches if you understand the historical context.

As a former Pixel and current Samsung user, I can relate to almost everything on this list except for the Camera app. What is it that you prefer about the Samsung iteration? I found my preference to be for the Pixel version.

It sounds like fundamentally SAP and Palantir target different use cases though? While SAP has OLAP functions, their bread and butter is highly domain-specific and transactional.

IANAL, but I assume this could open Wikimedia leadership to charges of contempt and eventually lead to needing to avoid visiting the UK or other extraditing countries and potentially pave the way for asset seizures. You generally don't want to antagonize world power governments.

Well, their huge GPU clusters have "insane VRAM". Once you can actually load the model without offloading, inference isn't all that computationally expensive for the most part.

When I get dressed in the morning, it's 58 degrees outside. Today there's a high of ~88. It's totally normal to look at the weather to determine what to wear.

This sounds like Gemini Code Assist rebranded under the successful Gemini CLI banner. I'm sure this was done to "consolidate" offerings and brands, but this is just way more confusing. CLI has a meaning, and this doesn't seem to have a CLI at all? Product looks cool, but the naming is just baffling

Was the post edited? It looks like the submission title is exactly the blog post title except with "Google Says" appended.

Yeah, perhaps sending fake invoices to customers? There's a lot of precedent for that:

https://krebsonsecurity.com/2025/07/phishers-target-aviation...

Interesting! This feels like they're trying to position it as a competitor to Nvidia's Omniverse, which is based on the Universal Scene Descriptor format as the backbone. I wonder what format world objects can be ingested into Genie in - e.g. for the manufacturing use cases mentioned.

This rings true. I feel this way about the incognito "tracking" lawsuit as well.

I'm pretty sure their endgame right now is just "don't lose all of your market share to companies that _will_ use LLMs".

I interpret it more as "maybe consciousness is not meaningfully different than sophisticated token generation."

In a way it's a reframing of the timeless philosophical debate around determinism vs free will.

I assume that an "advanced version" of Gemini Deepthink means it was a different model or had significantly more test time compute than the upcoming Deepthink in the Gemini Ultra subscription. Still cool to see OpenAI and Google neck and neck.

You and the parent poster seem to be conflating the ideas of:

- Does not have the requisite skills and experiences to do X successfully

- Inherently does not have the capacity to do X

I think the former is a reasonable standard to apply in this context. I'd definitely say I would be bad if I tried to play the guitar, but I'm not inherently incapable of doing it. It's just not very useful to say "I could be good at it if I put 1000 hours of practice in."

I have no problem taking issue with using the term "engineering" for this, but this reads like OP does not believe that you can get better results based on the way you structure your prompt and what information you include. That is wild to me and should be obvious to anyone who spends more than a few hours using LLMs.

For what it's worth, I think about the "engineering" in "Prompt/Context Engineering" almost more akin to how it's used in "Social Engineering". You are influencing the model to produce a desirable result.

Yes, but this is precisely the point isn't it? It's blatantly enabling and embracing "othering" for no technical reason as an explicit strategy to exploit social pressures to maximize profit.

I'm reading through the Expanse books and a passage that I read after seeing your comment jumped out at me. It's a quote from Havelock:

"As a boy living planetside, he had always thought of space as cold. And while that was technically true, mostly it was a vacuum. And so a ship, mostly, was a thermos. The heat from their bodies and systems would bleed into the void off over years and decades of it had the chance."

So at least the books got it right!

I still remember sitting through a legitimate field medic first aid course before unlocking the medic class. That game was something else!

The AI Overviews (on the main SRP) is pretty hit or miss. The new "AI Mode" (separate tab) is _very_ good.

Why do you hate Cloudflare so passionately?

In the immortal words of The Dude: "You're not wrong, Walter, you're just an asshole."

Interesting timing with the supposed Robotaxi launch later this month

Thank you for this. One of these links should really be the one in the submission. The Mozilla petition doesn't provide any useful context.

Almost all of those benchmarks are coding related. It looks like SWE-Bench is the only one where Claude is higher. Hard to say which benchmark is most representative of actual work. The community seems to like Aider Polyglot from what I've seen

> Few people feel the need to learn it.

Is this true? Almost every educated adult under age ~45 that I encounter touch types. Certainly in the white-collar/tech professional sphere.

There is definitely a generational divide where it becomes far less common with those who did not grow up with ubiquitous computer usage.