Yes, anything, not just software. I wonder if monopolies will somehow start falling or if they will try to defend themselves, are we on a path of continuous decentralization?
I wonder how many chat logs they did use in model's training data? If model consumes human’s reasoning (the chat logs) and then presents results as its own, it's a kind of parasite right?
Harvesting collective output of human consciousness to build AI that then replaces the very people who provided the input.
Replacement isn't bad as long as it doesn't strip people of dignity, and that's the tricky part.
Google, in a large sense, replaced librarians, service desks, 411, and many other specialist people and services whose purpose was "The knowledge is somewhere, and there is an index, but most people don't know where the index is or how to read it."
But librarians aren't really diminished as a result (while service desks may have kind of quietly evaporated to go do something else).
If the system becomes "People talk about mathematics and then an analysis engine reads that zeitgeist much faster than people can and pulls it all together into new insights," that's not apriori bad... Except that we have to consider how many people were in that conversation because their ultimate goal is to have an insight tied directly to them, i.e. to become the Hamilton in "Hamiltonian operator." Because the reality is that people practice mathematics both to advance the sum total of human knowledge and to roll the dice on getting stamped into the history books as the closest form of immortality that people get, and we can't just ignore both drives.
Grant Sanderson (3blue1brown) expressed thoughts on this topic space in a recent video, where he suggested one possible shift might be from the focus in the field being on frontier-pushing to the focus being on making the frontier more "human-shaped." There's a huge gap between higher math and what most people understand, and possibly ample opportunity to bridge it.
Aggressive marketing (including daily posts on HN). Many people simply don't know about alternatives, there are many people who never heard about pi and opencode and live comfortably in claude-codex bubbles.
I find it curious that almost every AI post has way more comments than other posts, and I sometimes wonder how big of a part agents play here. I can't seem to distinguish whether it's just real interest in the subject or a big campaign trying to influence people's opinion
It's on top of mind for everyone in the software development industry (be it SWEs, product people, designers, managers...), as it's such a "cross-cutting" concern, I am not surprised it's always on the HN front page.
Sure, some Rust 1.98.7-beta release is interesting to some, but AI affects most of us, one way or another.
I (and many other people I personally know, so they are not botfarms) feel many different strong emotions regarding AI on a daily basis: anxious, frustrated, tired, bored, suprised, entertained, empowered, optimistic, pessimistic, usually all of it almost every single day.
Obligatory PSA that OpenCode does not respect prompt caching, OpenCode will cost much more whenever there is a price difference between cached and uncached input tokens. It mutates the system prompt (at the very beginning of the transcript) every single request with (among other things) the current date and the current working directory (which changes if the agent ever runs `cd`) and will repeatedly invalidate the entire cache.
> Mathematics has very little to do with the real world
But it's a construction based on very basic properties of the real world, at least on those properties we accept and perceive as very general and we use them as building blocks.
When one player tries to monopolize AI tech and market by all means, others are those who don't want to be hooked to a foreign will in the future. I think this is the reason in doing open research, and it's likely positive for all of us.
Yes, it happens quite often recently. I think, their approach is "dominate and tolerate" in a controllable way, so they know quite a bit about who is actually using NewPipe, yt-dlp, etc. and they tolerate it in order to have a kind of leverage to preserve their monopoly. Then if number of out-of-official-app views becomes like above certain point, they aggressively block with "bot detected" message.
Compaction is painful if you run just one local LLM, the best way to avoid it is to keep context as small as possible.
One trick I find useful is to have one model with two KV caches running and while first cache has produced tokens, second cache immediately summarizes them during input tokens are being generated (tools time), then harness switches to the second KV cache which takes newly produced input tokens while KV in first cache is getting replaced with compacted summary tokens. This is a kind of ping pong, so we trade more space for less time. Still experimenting but it looks it works, and nice bonus it improves GPU utilization. Btw I have my own harness and model serving code, but it can be easily implemented in any other harness and model server.
I tend to ask a line of questions to the LLM as first step to a point the context is enough for me. Then i ask for the work i need but then get back to previous context using /tree.
It's like asking the questions needed to get the answer you need from the LLM, can be either an answer for a plan/todo or a task to code/change something.
reply