pull down to refresh
0 sats \ 0 replies \ @ajonas 13 May \ parent \ on: What's something people only romanticize because they've never actually done it? AskSN
The more things change, the more they stay the same.
Thanks for re-surfacing @elvismercury. As mentioned in #396527, this is a Bitcoin Dev Project project and the source code is here -- contributors and PRs welcome!
Nothing is being trained here. The responses are generated based on the same sources as bitcoinsearch.xyz. We take the content from the most relevant sources and then ask OpenAI (could use another LLM) to summarize the raw content. This limits the hallucinations as we don't make up any answers that don't have content to support it, but also limits its range. It does well with technical questions that people have written long mailing list posts about, but not very well with general purpose bitcoin things. We also implemented L402 for DoS protection.
This is not meant for someone new to code. You are welcome to try but that’s not who we are targeting.
This is different than the seminars, which I think were a solid educational experience but didn’t prompt many to take the leap to being a ₿OSS contributor. This program is singularly focused on that outcome.
https://chat.bitcoinsearch.xyz/ uses it. I’m not aware of many others that have implemented it.
Yes, we allow a certain number of questions for free.
No plans on turning this into a business. The payments collected are paid out as sats-for-review on review.btctranscripts.com.
This is a Bitcoin Dev Project product.
The responses are generated based on the same sources as bitcoinsearch.xyz. We made some choices to cut down on hallucinations, which limits its range and so it thrives with deeply technical questions and won't do very well with general purpose bitcoin things. We've also implemented L402 for DoS protection, which I haven't seen much of in the wild.
I’d steer you to LibrerÃa de Satoshi to find technical, Spanish focused programming.
@elvismercury Here is another project relevant to this discussion.
In my experience, what you can do, oftentimes, is remove auxilliary frictions.
Absolutely. This is a particular area where bitcoin gets its lunch eaten by ethereum, et al. My favorite tool, polar, is beautifully done, but its not Jamal's full-time job or anything. We've explored streamlining dev environment setups, but it lacks polish.
Warnet and SimLN were two other attempts at adding to tools to be proud of list, but we aren't there yet.
Lots of work to do.
Thanks @elvismercury. The idea of "bundles" is a useful insight.
To riff on that, I think the "bundle" can also be quite personal -- different learning styles, experience levels, and time availability. I agree that there remains a lot of room for improvement, but simulating these activities can only go so far. To do the work, future contributors must take the step of doing the work itself. That sounds like a tautology, but it remains the biggest obstacle that I've observed separating learners from contributors. The former is passive and safe, the latter is active and scary. This program is designed to break down the barrier, but ultimately the participant is going to need to take their own step off the ledge.
The tension between simulating the work in order to help build confidence (i.e. toy projects and tutorials) versus doing the work itself at a remedial level (review clubs and good first issues) leaks its way into the medium itself:
We've made videos, podcasts, onboarding guides and workshops. We have cute/approachable ways to ask technical questions like ChatBTC and a coding game (WIP). Some of these are passive and others more active, but none of these are doing the work itself.
There exist networks of technical meetups and learning communities (i.e. Chaincode seminars, Qala, Library of Satoshi, Bitshala.
Each of these learning communities have their own flavor, but many participants hide in the comfortable position of being passive students in lectures. The residency has been the most effective bitcoin onboarding program I've ever been involved with because we cut the field down to 3% of applicants and required them to quit their jobs, thereby, burning the bridge behind them. There was no going back and so they could only go forward. The residency model is effective, but is not the accessible and scalable onboarding that we need.
Action remains an elusive outcome for many of those learning from the current bundle. My hope is that this program can be the catalyst to bring scores of people to where they want to go.