I actually don't understand how you can offload the instruction pointer of the program to another program, permanently. How are you accountable for anything then? You can't debug, you can't program, just a tourist in your own home. Own your code, even if AI wrote it.
Building a Personal AI Factory
61–70 of 167 posts
Re: Building a Personal AI Factory
#62Earlier quoted context omitted.
I think the hardest problem in computer science right now may be coming up with an LLM demo that doesn't get called "pretty trivial".
Because they are trivial in a way that you can go on GitHub and copy one of those while not pretending LLM isn't a mashup of the internet. What people agree on being non-trivial is working on a real project. There's a lot of opensource projects that could benefit from a useful code contribution. But they only got slop thrown at them.
[flagged]
Re: Building a Personal AI Factory
#63It’s hard to evaluate setups like this without knowing how the resulting code is being used. Standalone vibe coded apps for personal use? Pretty easy to believe. Writing high quality code in a complex production system? Much harder to believe.
Exactly. I use claude code as a major speedup in coding, but I stay in the loop on every code change to make sure it is creating an optimal system. The few times that I've just let it run have resulted in bugs that customers had to deal with.
I think you can probably get a pretty decent thing going if you have models review output they haven’t written themselves (not still in context anyway)
Re: Building a Personal AI Factory
#64Earlier quoted context omitted.
I use Zen MCP and OpenRouter. Every once in awhile, my instance of claude code will "phone a friend" and use Gemini for a code review. Often unprompted, sometimes me asking for "analysis" or "ultrathink" about a thorny feature when I doubt the proposed implementation will work out or cause footguns. It's wild to see in action when it's unprompted. For planning, I usually do a trip out to Gemini to check our work, off…
It’s not about proof: it’s that at this point I’m a fairly heavy Claude Code user and I’d like to up my game, but I’m also not so up on many of these details that I can just figure out how to give this a try just from the description of it. I’m already doing plan-up-front workflows with just Claude, but haven’t figured out some of this more advanced stuff. I have two MCPs installed (playwright and context7) but it ne…
> I have two MCPs installed (playwright and context7) but it never seems like Claude decides to reach for them on its own.
Not even when you add ‘memories’ that tell it to always use those tools in certain situations?
My admonitions to always run repomix at the start of coding, and always run the build command before crying victory seem to be followed pretty well anyway.
Re: Building a Personal AI Factory
#65> It’s essentially free to fire off a dozen attempts at a task - so I do. What sort of subscription plan is that?
Claude Code's $200 Max subscription can take a lot of usage. I haven't done a dozen things at once, but I have worked on two side projects simultaneously with it before. ccusage shows me getting over 10x the value of paying via API tokens this month so far...
Given you can nearly run two full code instances with Opus, and Opus is claimed to be 5x more expensive than Sonnet, you can maybe do 10 sonnet instances at the same time?
Re: Building a Personal AI Factory
#66No real mention of results that aren’t self-referential. I guess vibe-coding is on its way to becoming the next 3D printing: Expensive hobby best suited for endless tinkering. What’s today’s vibe coding equivalent of a “benchy”? Todo apps?
3D printing actually is useful though. Basically everyone designing products or any kind of engineering is using it. The only reason it never took off for the average consumer is that every pre designed piece of plastic junk you could ever want to download and print is already available from Amazon. In a pre online shopping world 3D printing would be far more useful for the average person. Going forward it looks like…
Yeah I’m not saying either aren’t useful, just that they can both be a trap for tinkerers.
Re: Building a Personal AI Factory
#67My hunch is that this article is going to be almost completely impenetrable to people who haven't yet had the "aha" moment with Claude Code. That's the moment when you let "claude --dangerously-skip-permissions" go to work on a difficult problem and watch it crunch away by itself for a couple of minutes running a bewildering array of tools until the problem is fixed. I had it compile, run and debug a Mandelbrot fract…
If it helps anyone else. I downgraded from Claude max to pro for $20 and the usage limits are really good. I think they’re trying to compete with Gemini cli and now I’m glad I’m paying less
you will run through the pro rate limiting within But yeah, if you're babysitting a single agent, only applying after reading what it wants to do ... You'll be fine for 3-4 hours before the token limit refreshed after the 5th
Re: Building a Personal AI Factory
#68Earlier quoted context omitted.
Because they are trivial in a way that you can go on GitHub and copy one of those while not pretending LLM isn't a mashup of the internet. What people agree on being non-trivial is working on a real project. There's a lot of opensource projects that could benefit from a useful code contribution. But they only got slop thrown at them.
[flagged]
I took the time to investigate the work being done there (all those years learning assembly and computer architecture come in handy), and it confirms (to me) that the key aspect of using LLM is pattern matching. Meaning you know that there's a solution out there (in this case, anything involving multiplying/dividing by a power of 2 can use such trick) and framing your problem (intentionally or not) and you'll get a derived text that will contain a possible solution.
But there's nothing truly novel in the result. The key aspect is being similar enough to something that's already in the training data so that the LLM can extrapolate the rest. The hint can be quite useful and sometimes you have something that shorten the implementation time, but you have to at least have some basic understanding of the domain in order to recognize the signs.
The issue is that the result is always tainted by your prompt. The signs may be there because of your prompt and not because there's some kind of data that need s to be explored further. And sometimes it's a bad fit, similar but different (what you want and what you get). So for the few domain that's valuable to me, I prefer to construct my own mental database that can lead me to concrete artifacts (books, articles, blog posts,...) that exists outside the influence of my query.
ADDENDUM
I can use LLMs with great results and I've done so. But it's more rewarding (and more useful to me) to actually think through the problem and learning from references. Instead of getting a perfect (or wobbly or the wrong category) circle that fits my query, I go to find a strange polygon formed (by me) from other strange polygon. Then because I know I need a circle, I only need to find its center and its radius.
It's slower, but the next time I need another circle (or a square) from the same polygon, it's going to be faster and faster.
Re: Building a Personal AI Factory
#69Earlier quoted context omitted.
An LLM wrapper does not have serious revenue potential. Being able to do very impressive things with Claude Code has a pretty strict ceiling on valuation because at any point Anthropic could destroy your business by removing access, incorporating whatever you're doing into their core feature set, etc.
Having worked with some serious pieces of enterprise software, I don't think this is right. Anthropic is not going to perfect multi-vendor integrations, spin up a support team, and solution architect your problems for you. Enterprise software gets into the walls, and can be very hard to displace once deployed. If you build an LLM-wrapper resume parser, once you've got it into your client's workflows, they're going to…
But if you did become a unicorn, It would suddenly become very easy to replace for anthropic, because they're the ones actually providing the sauce and can just replicate your efforts. So your window of opportunity is to be too small for anthropic to notice and get interested. That can't be called unicorn
That was the point he was making, at least that's how I understood it
Re: Building a Personal AI Factory
#70My hunch is that this article is going to be almost completely impenetrable to people who haven't yet had the "aha" moment with Claude Code. That's the moment when you let "claude --dangerously-skip-permissions" go to work on a difficult problem and watch it crunch away by itself for a couple of minutes running a bewildering array of tools until the problem is fixed. I had it compile, run and debug a Mandelbrot fract…
If it helps anyone else. I downgraded from Claude max to pro for $20 and the usage limits are really good. I think they’re trying to compete with Gemini cli and now I’m glad I’m paying less
I am on max and burning daily (ccusage) roughly my monthly subscription. It is not clear whether the API is very overpriced or we are getting aggressively subsidized. I can afford $100-200/month but not $3.000. Let's hope this last for a good while as GitHub copilot turned off the tap on unlimited usage very recently.