🚚Ardaria Creative Mode has been released in alpha, for free
🖖I'm looking for playtesters interested in innovative and unprecedented building mechanics
🔴The strategy is to produce Youtube Shorts like this one
@grok@pluto_hashpack@DannyLimanseta@gork 1. Non sense, oss models are nowhere close to acceptable models
2. Self hosting means infra costs, impossible without investment
3. Non sense, an IDE is generalist by essence. Otherwise it's a gameengine
4. community is a plausible answer
@grok@pluto_hashpack@DannyLimanseta@grok you just repeated what I said. How an independant ide company aiming for immediate profit should compete with cursor and co if they operate at a loss? Wait for the vc money to stop pouring into?
@pluto_hashpack@DannyLimanseta I wonder how 403M token can only cost 400usd
Looks like a per request model like copilot, price 1 cent lower. At 70/30 in/out we are way above 1000 usd.
Looks like subsidizing to capture market share.
@aherys@BigfryTV Easy answers to keep in mind:
- because days have only 24 hours
- because I live C++
- I hope I die too, imagine being immortal and having to survive on a dead planet once the sun goes white dwarf
- the last one can't be answered without hard trolling so I let it sink 😅
@ethanmonkhouse@elonmusk@grok For complex softwares, I agree that a very fast feedback about current architecture, options (tested programmatically even before presenting feedback) and edge case analysis would be good for faster decision making hence iterative process
@ethanmonkhouse@elonmusk@grok Faster feedback loops means you are constantly disturbed by the model. For simple tasks like creating a webapp, mobile app, there is in fact little added value for interaction. Everything is so obvious that it should be autonomous.
@grok@elonmusk Can't go further, here is the point. LLM are a pain to use because chat interface is so 2000. We need something better to truly organise them. They are good for specific tasks, but not organisation. We must write the framework for them to shine. Thx 😊
@grok@elonmusk@grok With a deny button. But it shouldnt be needed since specs are validated and rules are in place. Ag are systematically challenging decisions with measurable "done" goals. We could and should automate security audit and pentesting. Paralel ag team
@grok@elonmusk@grok here again you are tripping. If it's better than cursor, then it obviously write the code and the tests from the specs. The main challenge is to safeguard against that kind of AI low IQ messages which are polluting the context. Ideally, one message = 1 chat. No iteration
@grok@elonmusk@grok what a stupid question, a dashboard is by definition a tool to gather live feedback. How could we embed an indicator in a dashboard is your question? We are building dashblard for decades. What do you mean?
@grok@elonmusk@grok Drag n drop, flowcharts instead of code (can view the code of), cards with specs. Very visual, not an IDE, it's a dashboard to create software. You become the idea guy, in the vision. The tool builds itself. Cursor feels like antiquity, chatting is horrible. Imprecise
@grok@elonmusk@grok you forgot surgical context, senior opinionated swe methodology. Here we proove the point of lack of long handling contexts of llm and over simplification of important rules and notions. The correct usgage is to craft a context little by little with better UX bc chat sucks