> Marble makes tools visible and easy to select for each task
Isn't that what current agents already do? e.g. codex: https://ibb.co/Y71qJJZb
6,090 karma · joined July 31, 2013
Isn't that what current agents already do? e.g. codex: https://ibb.co/Y71qJJZb
I also did some light simulations so we could see how the skirting roof around the building would allow light in for summer vs winter.
We have been very happy we used VR. It allowed us to iterate on what felt right, communicate exactly what we wanted to the builder, and the sunlight simulations have allowed it to be comfortably usable year round - letting light in during the winter, but keeping direct light out during the summer.
Dont try and handle the entire project in context.
Use a well structured filesystem layout for your code with a few lines in an AGENTS.md describing the layout and core architectural requirements (no more than that, as per the article!).
Then work on small-medium tasks at a time with a fresh context.
At the end of your task, ask the agent if there are any key points about the project layout or architecture it would want to add to memory - audit those manually and amend your AGENTS.md accordingly.
If your code is well structured and you keep your tasks localised, you can get away with seemingly minuscule context windows.
It's also worth noting that high effort models love to slurp up whatever context they can. You almost never need/want high effort for non cross-cutting tasks.
From my experience this holds true to this day. It was one of my core observations for similarity to the limitations of human working memory on "Engineering for Bounded Cognition"
By analysing a specific piece of prose you probably arent using the same chunk as would otherwise be analysed and can end up with a different result.
Not really the same. Private citizens have received fines MUCH higher per-work when downloading for just their private consumption.
Personally, I dont like theses digital "purchase" licenses at all. The only "rights" DRM is used for is to restrict those of the consumer - a far cry from the transferable digital rights that big-media touted that resulted in them receiving overreaching protections for copyright in the first place.
Unless they are transferable, all digital licenses should be short term (<= 1mo) and marked clearly as a rental to avoid consumer confusion.
"Next, open Xcode"
You are gatekeeping based on the medium, not the skill or motivation.
Your prejudice/ableism in calling it "sloppyfying" is exactly the problem I mention with allowing it to be a flag.
Some people already dismiss genuinely useful content solely based on the use of AI to assist in writing it - i am not sure what flagging would do other than to reinforce that prejudice.
I posted a couple of my articles here, and the one that got traction was generally well received (and also received some constructive feedback from those who acknowledged that is was AI assisted) - but it is evident across HN there is a vocal minority who outright dismiss content solely because it was "AI generated" completely disregarding the content itself. I appreciate this is personal taste, or LLM fatigue, or whatever, but its not really constructive.
If what you want to do is target the slop while not targeting the quality content, then that is what the voting mechanism already does. If people don't like something they can downvote it. Flagging content as AI generated is just a dogwhistle to those who want to downvote AI generated content. If anything, id rather see a rule that stops people commenting on stuff just to dismiss it as "LLM slop".
Its already trivial to avoid detection with fine-tuned humanisers [i] built on non-instruction-tuned models. That makes the flag mostly useless - or worse - a way of penalising any content you disagree with. I'd rather not hide what I am doing and have something that I feel reads well than hide it and sacrifice the message to satisfy a vocal minority.
[i] https://arxiv.org/abs/2605.19516
EDIT: downvotes, as expected. hope you see this anyway @dang. Downvotes kind of make my point for me.
Made a talking head with some idle animation and visemes and some broken crt-like effects. The meat of it is only a few hundred kB - i can probably make it even smaller with making the graphics smaller.
A bit of post processing on some narration for extracting mouth shapes and it seems to work quite nice as a low-footprint retro talking head. Im thinking i'll make it some kind of chatbot interface.
Its very much a WIP, please don't be too critical - i am only sharing because it is fun :)
How is that counter-intuitive? You changed the key, so it didn't return a value. Thats a KV cache working as intended.
I sold my switch sports a few months after getting it for this reason. We still play Wii sports (and resort) on the Wii-U.
Thats not true at all. You can use the Agent SDK [1], which uses your subscription [2]. I use it via ACP [3] with custom system prompts and tooling. I have found it very powerful and flexible. It has its own agent loop, of course, so maybe thats the limitation using it with opencode?
[1] https://code.claude.com/docs/en/agent-sdk/overview
[2] They were talking about giving credits for the SDK usage rather than it using your allowance directly, but that seems to have been put on hold for now. If and when that changes, I will likely jump ship, but I am more than happy with it right now.
[3] there isn't an official ACP wrapper - zed have one but its quite limited. its trivial to build one though, or you can just use the SDK directly and wire it into your interface of choice.
You can reduce a lot of the token use for traversal by giving your agent access to some form of LSP in addition to hierarchical direction with your AGENTS.md (or equivalent) for monorepos - but a spread-out codebase is always going to end up requiring some form of traversal to solve each task.
And that traversal isn't just token use - its repeated round trip latency (LLM (queue time -> prefill -> decode -> output) -> Agent (parsing -> tool call -> tool response) -> back to LLM) for EACH step (well, some can be done in parallel, but in practice its mostly sequential) - slowing down the task considerably.
Locality and structure are key when it comes to efficient use of agents. The context window is always bounded and attention across it is inconsistent.
I would please urge you to read further into the manifesto itself but would also recommend you start at the foreword so you can understand the reason for the use of AI assistance in my writing.
The actual criticisms you have about the content however, I'd like to challenge:
The "adding up to one" is just a simplified gloss over softmax. It's very possible it reads poorly, and thats on me - not LLM gibberish.
As for the incoherence - I have to totally disagree. You have merged the 2 things the post keeps apart - capacity and attention over it. That a model can swallow a schema and write code is a competence humans share. We have been doing it for decades. Besides, the claim was never about us sharing capacity (other than stating it is always bounded) - it was about our attention failing in eerily similar ways.
So, AI slop, no. AI assisted, absolutely. It's sad that some judge the "who" more important than the "what" - especially for this kind of writing. But it's fair feedback nonetheless. I'll see what else I can do with assisting my delivery.
They literally tried to kill it - stating (nonsensical) reasons why it was obsolete and unneeded.
And since now the rest of the world have adopted it despite Google, they have crawled out of their slime pits praising themselves for its development with only a passing mention of cloudinary?
Sickening.