lots of highfalutin language trying to make something thats pretty hand wavy look like it's not. Where are the benchmarks? The "vector algebra" framing with @X + @Y - @Z is a falsehood. Embedding spaces don't form any meaningful algebraic structure (ring, field, etc.) over semantic concepts, you're just getting lucky by residual effects.
I think a missing key detail that becomes obvious in hindsight is that you needed a lot more structured logic before when llms were not as powerful, but now that these models are being post trained on this type of minimal tool breadth but high depth (ultimately towards just an agent using a computer) you can remove a lot of the scaffolding.
We are almost at state of being able to make the hedgefund name generator for the dozens of places doing same thing. Combinations of flow and whatever suffix {ly, drop, wise...} and boom its probably one of these wrappers.
| Kinda surprised they didn’t run into model collapse problems
Not sure why you would expect this, all the models started doing this as its much more cost effective to get data for post training don't you remember the first grok release where many times it started replies "as a model trained by openai..."
The quality is most definitely suspect for how much revenue it brings it and how poorly its allocated. MTA has been full of cronyism and corruption for years and the cycle of kickbacks. Yes, its a complicated service, however you cannot deny the lack of transparency and ineptitude leaves the service in a much worse place than it currently could be. People can understand price increases when it translates to service.
I've rarely seen any non elementary use cases where just giving access to an MCP server just works, often times you need to update prompts to guide agents in system prompts or updated instructions. Unless you are primarily using MCP for remote environments (coding etc or to a persons desktop) the uses of it over normal tool calling doesn't seem to scale with complexity.
Would think that blue origin and project kuiper launching for amazon that would put downward pressure on SpaceX, as they are about to have a huge amount of competition for starlink, as Amazon has massive distribution advantages - wouldn't be surprised introductory bundling with Prime etc...
I mean - social engineering of humans takes many forms that can definitely include linguistics of persuasion etc... But the core thing to me fundamentally remains the same is the LLMs do not have symbolic reasoning, its just next token prediction, guardrails are implemented via repetition in fine tuning from manually curated examples, it does not have fundamental or internal structural reasoning understanding of "just dont do this"
based on linkedin it looks like they took a failed company thats existed for a year "General agents" which had what just looked like a single browser agent and now have rebranded that.
the need for more web search indices is indeed dire given landscape with agents and providers turning into walled gardens means that independent ones are definitely going to be needed, but just seems insurmountable when building actual index is so costly. Maybe just purely pareto efficient of serving 80% of requests or something is good enough.
seems like you are misappropriating what canaries are useful and used for... they are designed to be lightweight and shallow... hence the name and whole analogy, canaries never were meant to determine if a mine was structurally unsafe etc
Montessori is just an educational framework, I have no idea where you draw broad conclusion that the one or two things you looked at deemed it be "rigid" or little opportunity... Sounds like a random bad apple. There's a correlation between gifted children and montessori because it allows them to develop at their own pace which is often faster than that of traditional classrooms etc, it's not for everyone.
The premise of this article is almost entirely wrong. The primary driver of overseas manufacturing was that labor and manufacturing costs in the US are higher. The part about USD is misconstrued, countries hold USD because of its stability, which we benefit from being able to essentially inflation as they all still have their own banks as well to regulate their monetary policy.
not really sure how that differentiates since those things you mentioned are ancillary to main value. Also - browser base is insanely cheaper, but looking at the prices this doesn't look like a real company mainly just a way to have users in free tier (with toy level limits)
you're missing that open ai and anthropic are finetuning their models for their coding agents, i.e. using rl and annotating datasets and optimizing for them with directly learnable parameters. Cursor is just able to use whatever foundational model apis are available (not sure if or when labs might give programmatic access to coding agents). I'm sure they are trying to train OSS models but those fall way short performance wise of proprietary ones.
frontier labs do finetuning of their models for software dev using the terminal/cli driven style, annotating datasets to solve programming in this fashion, and fine tuning will almost always make for better performance. Cursor as mostly a wrapper is just using the underlying foundation models in their framework and orchestrating on top of that, as opposed to doing actual learnable objectives in training to make things better.
not only that but the way that openai and claude have their own foundational models/agents trained to work via CLI, which will basically always be better than just cursors gpt wrapper approach.
The reason is abundantly clear. Cursor was just a GPT wrapper with a nice UI/UX (which was very nice when it came out) it has some other models like autocomplete as well, but its still a wrapper. OpenAI and Anthropic build and train models specifically to work via CLI driven processes, which is why they are so much better now. Cursor is basically dead as I'm sure they realized they get much better performance with the CLI/agentic approach.
Amtrak is not reasonable prices compared to trains in like any other part of the world, except some parts of UK. Other than booking extremely early you're looking at 150$ plus tickets.
the demo video is literally just single thread tool calling to external sources. Indexing data is also a really complex problem more than just adding some elastic search to gmail which also you will find does not scale easily, if that's even what you're doing.
that just sounds like you have no idea what MCP is, I don't even like MCPs but I can't even understand what angle you are coming from unless they specifically mean using external MCPs instead of your own, since it is you know open source...
what's interesting about this one is that their claims about what makes slashy different are almost entirely wrong... almost all the big models let you connect and do all of the things mentioned. Not understanding MCP at all is hilarious. If an agent has toosl to access multiple data sources it will make calls across them to resolve things, not sure whatever claiming but there's no way you are actually indexing at scale and probably doing just the exact same thing.
same, so much so that I almost think they have a bot army for astroturfing, because I cannot believe anyone gets value from it over just one of the main providers and google. They dont index data nearly as much or as often as google, they basically are at a pareto frontier of just serving the most amount of people with lean indexing based on fact that people look up a lot of the same things. It reminds me of looking at google earth where you can see cities have 3d models but a random neighborhood where you grew up would be images years old. So yeah, its useful but definitely not a replacement for google, so idk why anyone would pay for it.
fyi - your application page has a race condition in authn middleware that shows a blip of a dashboard layout without being logged in... For a payments platform that is a little iffy.