> The registry grows with use. Every session is smarter than the last.
This feels a bit like one of those “now you have two problems” solutions. After a few dozen sessions I would expect the tool registry to be full of “noise” for most prompts. I would also expect most tools to be extremely specific to the task at hand, leading to redundancy and ultimately poor programmability due to inconsistencies between tool APIs.
It's an open experiment, the utility of tendril is the concept. I am more curious about how good can the tool making get. Frontier models tend to be very specific about what they build so don't get specific bloat (yet).
> sample solutions from the model with certain temperature and truncation configurations, then fine-tune on those samples with standard supervised fine-tuning
It’s all moonspeak to me. I tried reading other comments that explain this and they all sounded different or contradictory. I’ve studied ML as a hobby years ago but this was before the LLM explosion. Guess I need to start over again?
In a few years when you're not 12 any more, you'll be embarassed by this. When that happens, don't sweat it, we were all 12 at some point. I'm just lucky that for me that was before the internet.
What are you trying to communicate in this comment? That you have spite for your users? Why? That you consider not bothering with Firefox support to be a good way to, what, express your spite? Do I have that right?
I support baseline browsers unless it’s not feasible otherwise. Sometimes things just aren’t possible in certain browsers. It’s expensive and difficult to design and implement things that fail gracefully. I’m not actually spiteful towards Firefox or its users; I _am_ spiteful toward other developers who feel they are entitled to leaving hostile comments for free hobby projects that don’t support their browser of choice for frankly technical reasons.
I was being facetious for rhetorical purposes. The OP I was replying to was unfairly hostile. I will also hazard a guess that they don’t have much experience writing sophisticated software for browsers.
I responded with the same sort of hostility to make my point that you’re not going to win “hearts and minds” for your cause by insulting developers for relying on browser standards that aren’t yet baseline. My point is that I am not persuaded by hostility, and I suspect other developers aren’t either. Users like this give their browser of choice a bad reputation when they make it part of their hostile identity.
At least investors like Amazon can afford to lose their investment ($50 billion). That would be like a normal person losing a few thousand dollars. It hurts, but life would go on.
That’s still $100B unaccounted for, and I’m pretty sure Amazon would expect fair treatment if other investors get a bailout. More likely, OpenAI is the one to receive the bailout, likely at the behest of the bigger investors, Amazon included.
That's why there's been such a massive effort to position LLMs as critical to national security. If they can make themselves big enough and critical enough (even in just perception) to the government they won't let them fail. They'll let individuals lose their livelihoods of course, since it's rugged individualism for all of us lowley normal people. But corporate socialism will keep the big players afloat.
This feels a bit like one of those “now you have two problems” solutions. After a few dozen sessions I would expect the tool registry to be full of “noise” for most prompts. I would also expect most tools to be extremely specific to the task at hand, leading to redundancy and ultimately poor programmability due to inconsistencies between tool APIs.
reply