Character.AI discovered the one conversation its bots couldn’t keep going: the one with its accountants. $56 million in compute costs in eight months, and the monetization strategy was apparently: hope the next investor is even lonelier than our users. Silicon Valley keeps confusing engagement with a business model.
If LLMs can’t be made more power efficient, then the technology isn’t ready yet for general use. There may have to be fundamental engineering breakthroughs before what these systems do can be more than a very expensive and inefficient worldwide proof of concept.
Of course, the military and the surveillance industry will still want to keep it, even if no one else can afford it, and we’ll be the ones paying for it.
They CAN be made more efficient. Hell, AMD just bought a company that translated LLMs to direct silicon… which should enable scalability if it wasn’t for one of the key players owning the tech thus limiting its use.
It’s also estimated that some 70% of Claude tokens gets spent on recursive self-correction so the output is higher quality. Instead of bettering the model, Anthropic just keeps throwing money at it to make the problem go away.
Meanwhile Chinese models - as small as 36-40-80b params - are outdoing these fancy frontier models if the right harness is used.
I’ve spent the last month or so using nothing but a local 27B Qwen model as my coding assistant and it’s working fine. The big American AI companies could all collapse in ruins tomorrow and it’s not going to slow things down all that much, anyone who wants to use AI will be able to continue using it. Might even be better in the long run if that happens - the people who are scared of ASI will get their much-demanded temporary slowdown at the frontier while everyone else catches up with open-weight models that everyone can use without central control.
Wanted to refactor a too large and hectic hacked together state machine into separate functions and reduce redundancy and increase code readability
Every model I’ve tried failed.
And that should be an easy task an actual AI should be able to handle.
Logic is done, just translate to a different structure.
That should be exactly what they are able to do at least.
None worked, because the project is just to big and interlinked.
I’ve got 24GB of VRAM, which helps a lot. I’m using the Q4_K_M version of this model, with the KV cache quantized to Q8, and that’s able to get me to 128000 tokens of context. I’m using the Qwen Code harness in VS Code.
If 128000 tokens still isn’t enough, break the task down into smaller chunks and help the harness out with MCP tools - I installed Codebase Memory to help it navigate large codebases more efficiently, for example. Ask the coding agent what’s causing problems and it may be able to help identify solutions too, I’ve had it troubleshoot itself several times now and it’s been quite good at inspecting its own settings and finding where the problems are with them. I even installed Firecrawl for it so it can do its own web searches.
When you add the additional codebase memory, that will also add all those mcp tools to each round-trip to the llm model. Do you still see improvements and less filling up if your context? Just curious how it works out; I can imagine that a text based memory that explains the architecture of the codebase would help as well. Not sure about the difference if using a graph to store the code context?
Depends on the framework you’re using. I’m using Qwen Code Companion (via a VS Code IDE plugin) and it’s got some kind of tiered system built into it for adding tool definitions to context, it’s got a standard built-in set that’s always mentioned but other stuff only gets added to the context under certain circumstances. I can’t recall all the details, I found out about it when I was chatting with Qwen Code to ask it why it hadn’t used a particular MCP I’d given it to solve a problem despite it seeming like a useful tool for the task and we got into a discussion about to configure it. As I recall it’s something along the lines of most tools only being present in context by default as little one-liner stubs and the agent has to tell the framework “that one looks like it might be useful for what I’m doing” for the full definition to get loaded.
Most of the time I’m working on small scripts and specialized applications these days, and codebase memory is overkill for those so I haven’t actually seen it in use much.
Character.AI discovered the one conversation its bots couldn’t keep going: the one with its accountants. $56 million in compute costs in eight months, and the monetization strategy was apparently: hope the next investor is even lonelier than our users. Silicon Valley keeps confusing engagement with a business model.
If LLMs can’t be made more power efficient, then the technology isn’t ready yet for general use. There may have to be fundamental engineering breakthroughs before what these systems do can be more than a very expensive and inefficient worldwide proof of concept.
Of course, the military and the surveillance industry will still want to keep it, even if no one else can afford it, and we’ll be the ones paying for it.
They CAN be made more efficient. Hell, AMD just bought a company that translated LLMs to direct silicon… which should enable scalability if it wasn’t for one of the key players owning the tech thus limiting its use.
It’s also estimated that some 70% of Claude tokens gets spent on recursive self-correction so the output is higher quality. Instead of bettering the model, Anthropic just keeps throwing money at it to make the problem go away.
Meanwhile Chinese models - as small as 36-40-80b params - are outdoing these fancy frontier models if the right harness is used.
I’ve spent the last month or so using nothing but a local 27B Qwen model as my coding assistant and it’s working fine. The big American AI companies could all collapse in ruins tomorrow and it’s not going to slow things down all that much, anyone who wants to use AI will be able to continue using it. Might even be better in the long run if that happens - the people who are scared of ASI will get their much-demanded temporary slowdown at the frontier while everyone else catches up with open-weight models that everyone can use without central control.
Pity about the economy in the meantime, I guess.
I’m always running into context limit problems
Wanted to refactor a too large and hectic hacked together state machine into separate functions and reduce redundancy and increase code readability
Every model I’ve tried failed.
And that should be an easy task an actual AI should be able to handle. Logic is done, just translate to a different structure.
That should be exactly what they are able to do at least.
None worked, because the project is just to big and interlinked.
I’ve got 24GB of VRAM, which helps a lot. I’m using the Q4_K_M version of this model, with the KV cache quantized to Q8, and that’s able to get me to 128000 tokens of context. I’m using the Qwen Code harness in VS Code.
If 128000 tokens still isn’t enough, break the task down into smaller chunks and help the harness out with MCP tools - I installed Codebase Memory to help it navigate large codebases more efficiently, for example. Ask the coding agent what’s causing problems and it may be able to help identify solutions too, I’ve had it troubleshoot itself several times now and it’s been quite good at inspecting its own settings and finding where the problems are with them. I even installed Firecrawl for it so it can do its own web searches.
@FaceDeer @eicker @floofloof @fonix232 @naeap
When you add the additional codebase memory, that will also add all those mcp tools to each round-trip to the llm model. Do you still see improvements and less filling up if your context? Just curious how it works out; I can imagine that a text based memory that explains the architecture of the codebase would help as well. Not sure about the difference if using a graph to store the code context?
Depends on the framework you’re using. I’m using Qwen Code Companion (via a VS Code IDE plugin) and it’s got some kind of tiered system built into it for adding tool definitions to context, it’s got a standard built-in set that’s always mentioned but other stuff only gets added to the context under certain circumstances. I can’t recall all the details, I found out about it when I was chatting with Qwen Code to ask it why it hadn’t used a particular MCP I’d given it to solve a problem despite it seeming like a useful tool for the task and we got into a discussion about to configure it. As I recall it’s something along the lines of most tools only being present in context by default as little one-liner stubs and the agent has to tell the framework “that one looks like it might be useful for what I’m doing” for the full definition to get loaded.
Most of the time I’m working on small scripts and specialized applications these days, and codebase memory is overkill for those so I haven’t actually seen it in use much.