Garlef

8 ms·

121.
by Garlef 6mo ago
I've taken the numbers from the alibaba pricing page which says $2-$6
122.
by Garlef 6mo ago
I'm not sure about this: An "infinite" canvas without some notion of recursion such as viewports feels incomplete.
123.
by Garlef 6mo ago
Agreed. The upside is that it does not leave the most important aspect open to interpretation. But it prevents this from being text-only at the point of creation: You'll most likely need some programmatic environment to create non-triv
124.
by Garlef 6mo ago
Opus 4.5 is already pretty good. Opus 4.5 is $25/m output tokens. This is at most $6/m output tokens. That's ~1/4 the price.
125.
by Garlef 7mo ago
https://www.youtube.com/watch?v=QWzLPn164w0 I don't like the hype language applied by the channel host one bit - and so this is not something where I expect someone tired of the hype to be swayed - but I think his pers
126.
by Garlef 7mo ago
“Everything has already been said, but not yet by everyone.” — Karl Valentin --- Personally, I'm still very interested in the topic. But since the tech is moving very fast, the discussion is just very very unevenly distributed: There&#
127.
by Garlef 7mo ago
Isn't this a common architecture in CQRS systems? Commands to go specific microservices with local state persisted in a small DB; queries go to a global aggregation system.
128.
by Garlef 7mo ago
> AI has also changed the dynamics around this. Splitting things into smaller components now has a dev advantage because the AI program better with smaller scope This is not AI specific and nothing new and also precisely why microservice
129.
by Garlef 7mo ago
But: Will it also let him win at Settlers of Catan?
130.
by Garlef 7mo ago
No problem: Just build a subterranean boat and launch a few nukes close to the core to restart rotation.
131.
by Garlef 7mo ago
"look ma, I've made the AI fail!"
132.
by Garlef 7mo ago
Maybe they should implement a graph based trust system: You need your favourite academic gatekeeper (= thesis advisor) to vouch for you in order to be allowed to upload. Then AI slop gets flagged and the shame spreads through the graph. And
133.
by Garlef 7mo ago
> anthropomorphism I think it's a topic worthy of discussion. But I would propably not leave it to Searle...
134.
by Garlef 7mo ago
serious question: > no change in survival rates > less series A would this not imply that companies got more efficient at using their seed funding? (But then again: The real dip in series A funding starts in 2018; so we might still se
135.
by Garlef 7mo ago
I think restrcicting this discussion to LLMs - as it is often done - misses the point: LLMs + harnesses can actually learn. That's why I think the term "system" as used in the paper is much better.
136.
by Garlef 7mo ago
Fullstack lean when?
137.
by Garlef 7mo ago
I like both worlds: Tinkering and vibe coding. My shift in perspective is really: Not all code deserves to be hand-crafted. Some stuff can be wonky as long as it does it's job. (And I think the wonkyness will reduce in vibe-coding as h
138.
by Garlef 7mo ago
I think this is 100% the right direction: Instead of imperatively letting the agents hammer your codebase into shape through a series of prompts, you declare your intent, observe the outcome and refine the spec. The agents then serve as a c
139.
by Garlef 7mo ago
Awesome! A few questions: - Is there a list of host languages? - Can it live in the browser? (= is JS one of the host languages?)
140.
by Garlef 7mo ago
A good rule of thumb: - Don't even let dev machines access the infra directly (unless you're super early in a greenfield project): No local deploys, no SSH. Everything should go through either the pipeline or tools. Why? - The mom
141.
by Garlef 7mo ago
> How many people can do that? The answer is simple: By definition only about 100-300 people. There's only 100 of the "worlds biggest companies" (assuming this refers to the top 100). And companies are usually started by 1
142.
by Garlef 7mo ago
I think you could also argue that LLMs in coding are actually just a novel approach at code reuse: At the microscopic level, they excel at replicating known patterns in a new context. (Many small dependencies can be avoided by letting the L
143.
by Garlef 7mo ago
> hallucinate the input True > fabricate the 'proofs' for the verification sandbox Not sure: It seems the tool here allows for non-llm code to perform validation inbetween the steps So in this sense, it allows you to manage
144.
by Garlef 7mo ago
> It’s hard to use external services I think it's interesting to add what they use it for and why its hard. What they use it for: - It's about automated testing against third party services. - It's not about replicating th
145.
by Garlef 7mo ago
I guess that's much harder: Those who measure correctly are likely also interested in saving.
146.
by Garlef 7mo ago
I think this is a very interesting question. But I think getting good AI features into a game requires a very very high engineering effort that is easily underestimated. Current game engines are very good at creating large worlds with inter
147.
by Garlef 7mo ago
I think this is the wrong mental model. Instead, we need better (self-explaining) translation from spec to code. And better tools that help us navigate codebases we've not written ourselves. For example, imagine a UI where you click on
148.
by Garlef 8mo ago
Maybe add something more based as a comparison? MSCI World or similar
149.
by Garlef 8mo ago
Could also just be a bias in the audience
150.
by Garlef 8mo ago
I what the article proposes is the wrong direction: Instead of every single person to maintain an offering a vertical slice through the whole stack, we should make it easier to publish content in the first place. The real issue that this pu

More ›