vineyardmike
It’s no secret that these big labs see “recursive self improvement” as a primary goal. It seems natural that the models would be explicitly trained to excel at all sorts of ML related tasks as a result. Is this one of those tasks, or is this emergent?
I wasn’t going to click on the click-bait headline without understanding the topic better, so the length of the original was unknown and irrelevant to me. The above comment tersely explains it enough to satisfy my prerequisite.
The comment, “slop” or not, was useful. When slop fills comments with junk and fake commentary, it’s far more annoying and detrimental than a basic summary. Slop-policing is exhausting to read - just downvote or flag trash instead of adding more noise that takes away from the actual content.
Is it not obvious that this is an Anthropic-made "personal-assistant-type harnesses"?
YouTube has had sponsors and inline ads for years from the content creators, so I’m fairly confident we’re already approaching the saturation point in some form.
It’s likely made no impact on iphone sales, but I don’t think it’s a reasonable goal. Android and iPhone users don’t meaningfully switch, the goal would be to gain Android customers, and it seems, drive market direction. 5% of all US phones, which is 10% of android phones, doesn’t seem terrible considering they’re certainly spending a lot less than Samsung et al in R&D
Android had a market share of 0, with no released products when they bought it. The 80%+ market share today was all Google.
Google Maps was purchased (pre Google IPO), but the original (“Where2”) was a desktop application. Google completed rebuilt it alongside the acquisition of other data sources, and released the web page. Years later, they built a mobile app which became one of the biggest apps of all time. Should we not attribute that success to Google instead of the Australian startup building a desktop app?
What is this question? Is the idea of new products, technology, or services strange and problematic to you? Convenience is a major driver of invention.
Also when I was growing up my parents would order delivery plenty - but the only option was pizza because the driver worked for the pizza joint. And yes, it’s true, their parents didn’t order delivery, but they also lived in crushing poverty of the Great Depression and could barely afford to feed their kids during every meal.
The website linked starts out with:
> We’re introducing GPT-6.1 Terra, a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing. A significant step up from GPT‑5.6 Terra, it excels at the communication and computer-based tasks that fill the workday, from drafting proposals and interpreting technical drawings to filling out forms and editing videos.
If you go father down, it has graphs and details with specifics of this, including:
> We measured GPT-6.1 Terra’s computer-use capabilities... At medium reasoning effort, GPT-6.1 Terra scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost. It even performs on par with GPT-6 Astra at only one-tenth its price per token.
and
> On MMMU-Pro... It even nearly matches the scores of our most capable frontier model, GPT-6 Astra, and GPT-6.1 Sol at much lower cost.
You can sign broken contract freely - it's unenforceable, so you can sign whatever paper HR hands to you, and only when you're retaliated against do you need to care. Which coincidentally is when you stop needing to be cooperative anyways.
Don't "YMMV" your way out of laws that protect you.
If your comp package has a non-disclosure clause it's illegal. You cannot be asked not to discuss salary in the US.
> People don't want to tip off
That's why it's illegal to be prevented from discussing salary. Discuss your coworkers - you might be the one being underpaid.
It's very common in these situations to target every org in the company with a goal of X% reduction. If a small fraction of these teams now need extra headcount, easy enough to allow those teams to hire. Usually you don't keep the severance if you return in under X days.
I've personally been laid off, only for the post-re-org team to hire several people, including asking to rehire me. It's rare, and those affected should probably plan to find new jobs, but it happens.
Massachusetts- 99% have healthcare, longer lifespan than most developed nations, top US education system, diverse economy, wealthiest population, strong safety net, high minimum wage, safe communities w low gun violence, net provider of federal tax revenue, etc
Alabama - best bet is to start driving north now
“No budget that runs at a deficit is acceptable”
Why not? Deficits can be healthy and safe when they’re managed. Literally most persons, businesses, and governments use debt. It helps manage purchases over time, and it’s an inflationary hedge.
The idea that “any deficit is bad” was literally pushed as propaganda to make you fatalistic about managing the deficit. As said, one party raises it, one party lowers it, and the party that raises it claims that lowering it to anything but zero is “just as bad” as their runaway debt growth. Propaganda.
My understanding is that most of the gains after the early GPT-3 "add more weights" era have come from post-training from synthetic data, especially through reinforcement learning.
I couldn't quickly find any public statements about the actual breakdown of training data, other than almost every lab saying that post-training size was "scaling" or growing "by an order of magnitude" from their previous major release.
If you bank with eg. Chase (biggest consumer bank in the US), you're not at real risk. In the hypothetical that the chase app is compromised, where does the money go? If the attacker moves the money to.. idk, US Bank, then US bank reverses the transaction and moves it out of the attacker's account. You can imagine hundreds of scenarios where the money is attempting to be moved to unregulated and irrecoverable destinations, but it doesn't exist at a scale that Chase et al. wouldn't be able to recover or accept the loss on to make customers whole (as is their responsibility).
You can't transfer "illions" to cash. You maybe can transfer it to a new bank - but then this same problem exists just one step away, with one more regulated institution to get involved. You can't suddenly fund/wire that money without a human calling you - if you try to wire even a much smaller amount, like <100k, as a random individual, then the bank will block the transfer and call you and ask about the money. If you're a corp that regularly banks in the "illions", you're not casually going to money launder. If you try to transfer it overseas, you're still in USD, so you're still in the regulated banking. Whoever exchanges that cash for a foreign currency is on the hook, and for "illions" of dollars, that entity is going to be regulated and behaving. This is all ignoring the fact that many banks and financial institutions will not accept an unexpected random large transfer (or thousands of smaller ones!) without some sort of audit or paper trail on the origin of the money.
As a mundane example, I recently bought a condo, and my bank required an in-person ID check to wire money over $25k, and then required me to later call and verify (twice) before I could start transferring my HOA payment to a random account. If you compromised my phone, that doesn't get you in-person with my ID in a branch, nor does that get you secondary confirmation over the phone (w/ a voice match check now!)
> which will disavow any responsibility as soon as they're on the hook for *illions
I haven't even touched on the political fallout if they disavowed responsibility, which is why all the regulations and KYC/CYA for the bank exists.*
Doesn’t make it a good product though.
Answering why you picked this particular path to high income is more interesting of an answer to the person asking and carries the conversation. Did your parents do it, did you have a professor that captured your attention, did you just google “easiest job to make a lot of money?” and pick the first one?
Someone asking wants to understand how and why you make decisions. What motives you. Etc. It’s the XY problem and from the other person’s perspective, you’re answering the wrong one.
(They support Matter protocol)
I’m honestly surprised in the age of vibe-coding we haven’t seen someone try to build out a direct “open walled garden”.
AI could solve the basic coverage on every operating system and screen shape, while the various p2p technologies have gotten really good and when combined with mDNS and maybe an encrypted byo-cloud-storage (eg drive, Dropbox, etc APIs) for async data synching. The nerd side is build a data sync layer, then a variety of apps and products can be built on top of it.
It doesn’t necessarily need to work as well as Apple, nor as rich features, it just needs to support enough devices well enough to allow heterogeneous-platform households or friends to share what they need.
When did we decide that isolation was freedom.
The software world is spoiled in that making something new is just text in a file, we’ve proliferated open source freeware, and with AI we can readily clone entire products in a weekend.
Sometimes though, people like to use commercial products for various reasons (reliability, cost, network effect, etc). If you like to tinker, this seems like a neat way for some without yak shaving an AI assistant along the way. We don’t need to reinvent every wheel for everything.
I personally think “embodied” assistants are a really fun idea, and this is just an API and SDK to build on and it’s super accessible to basically everyone in the world. If you were inclined to use Muse already, why not tinker with this. I’m sure the hardware would be easy enough to point to an API running on your own server later if you wanted.
I think it’s somewhat impressive they got it into orbit quickly since surely a year ago this wasn’t an earnest idea, and I think it’ll be interesting to see the affects of space on high end chips and connectivity fabric.
But also I agree that this just doesn’t make sense from a performance standpoint.
Also, I’d imagine the token-maxed user is a programmer that lives in chat. I’ll admit to having asked the LLM to move a method up/down in a file, and watched it burn tokens for a minute thinking and executing a menial task.
Many of these obligations are with the big clouds who don’t have capacity to serve them anyways. Google has said they have hundreds of billions in purchase obligations they can’t fill because they don’t have enough TPUs and data centers. Google gets a bigger marketing and investor headline, and there’s a queue behind Anthropic if they bail, so of course everyone would sign the commitments knowing the risks.
I’m sure AWS and Azure are on the same page.
In the worst case, many of their biggest customers have tons of GPUs (Meta, AWS), so they could always license out their raw models at steep discount to help absolve themselves of the obligations.
Finally, in the worst case scenario, even with margin compression, they pre-purchased most of the compute so there’s not enough for the open models to run on. If Opus was the same price as Kimi/GLM, I doubt many people would pick the open models.
The world looks pretty different today. Even if the model maker companies go out of business (I’m skeptical), the model weights would stick around (many are open freeware already).
I know many software engineers who haven’t written code this year. AI chatbots are regularly used as alternatives to searching manually by many people. Agents are becoming a valuable new enterprise tool, and now consumers tech consumers are hopping on board.