After they acquired Cursor, Grok 4.5 seems like a completely new model, performing at Opus 4.6 level, I'd say. But much cheaper and faster.
3,760 karma · joined May 10, 2020
LLMs are the best PDF-to-markdown converters, in my experience. I have a CLI that converts PDF to PNG, then run a background agent to "read" each PNG and write it down as markdown; it works flawlessly even for complex math formulas, it can "translate" complex charts, graphs, and tables into words.
It's slow and arguably expensive compared to traditional OCR, but very effective and precise.
A very plausible explanation for the adenoma detection rate to have gone down is simply that its prevalence went down among the population in the second three-month period.
This was not a randomized trial. Concluding that "AI usage degrades physicians' skills" is questionable at the very least.
The same applies to teaching "street smarts" to kids. You don't do it by throwing them in a hostile environment where they'll be prey to hostile people without having any defenses built up first.
[1] Homeschooled Children’s Social Skills: https://files.eric.ed.gov/fulltext/ED573486.pdf
[2] Homeschooling and the Question of Socialization Revisited: https://www.stetson.edu/artsci/psychology/media/medlin-socia...
Edit: if I had to bet (don't know any research), schools nowadays are the main producers of intolerance, with the indoctrination and teaching kids to only respect civil discourse, ideas and opinions if they agree with the mainstream world model.
And it's not a coincidence that their software is closed. They can command ridiculously high margins and continue to invest in high quality products.
I think we do need to build new interfaces, though. The existing ones were designed for humans to use, either GUIs for end-users, or APIs for developers. But LLMs have very different reasoning patterns. They even make mistakes in different ways.
What I've experienced in practice connecting LLMs to existing APIs is that LLMs fail miserably with the interfaces, but simply "translating" the interface in a way that is easier for them to "understand" solves the issue.
exactly! what I'm experiencing is that prompt engineering has its limitations and comes with inconsistency issues...
by designing the tool from scratch tailored to LLMs, we can make the interface match what their "own idea of how to do" that particular task, which is more reliable and scalable
While developing tools for LLMs, my team [1] and I came to the realization that we need a new engineering discipline. One that cares for the "machine experience", for building interfaces that are tailored to LLMs, having their 'preferences' and quirks in mind.
The LLM has to be seen as a consumer. A user itself.
We need a new breed of engineers dedicated to what we may call 'Machine Experience (MX) Engineering', just as we have UX Engineering, for instance.
Aren't you you the one setting up imaginary moral bars for what kinds of features a car should or should not have? And for what people should or shouldn't want in a car?
Aren't you the one believing "we're only allowed to combat human extinction" if we jump across your imaginary moral bars?
works pretty well, especially because you can use a more capable model for architecting and a cheaper one to code
Or because you spend precious sleep time on pointless comments online...
i can imagine this happening if a team has a myriad of hardware/os flavors and different server setups.
...even backed by crucial US supplies
some people think SWE is about "logic". it is, in part, but the "engineering" in software is much more of an art than it is in other branches, like construction
the current sota AI is great at logic and terrible in creativity and actual engineering. if the technical assessment is not designed for you to show your creative engineering side, do it yourself, do more than you were asked, think about what would be relevant to that company in terms of engineering creativity and offer that
that's the best way I know of showing you're a real engineer, not an LLM operator, it's worked well for me in the job search process
good luck!
started working with it this week for a new project
gosh, it's so painful and unintuitive... I find myself digging deep into their code multiple times a day to understand how I'm supposed to use their interfaces
the same foundation that makes the binary model of computation so reliable is what also makes it unsuitable to solving complex problems with any level of autonomy
in order to reach autonomy and handle complexity, the computational model foundation must accept errors
because the real world is not binary
one can always refuse to use