My guess is that this is going to be everything other technology that's democratized. You see a flood of low quality output because you have a lot of new non-technical devs. Some of these are good enough to crowd out some of the prexisting tools. The volume creates noise which also makes the good stuff harder to find. Eventually an ecosystem starts forming around these low hanging products which fill the gaps between pros and amatures (think of what happened to video editing and Apple). Eventually you have more people creating a better product in the long run. There is a bit of a feedback loop here as AI gets better, it makes the products it outputs better, which inturn can benefit AI as it learns from improvements.
I wonder if we'll reach a breaking point with public forges, where they'll simply reject hosting a repo if it isn't from someone with a vetted background or if it detects hallmarks of LLM slop (e.g., many commits over a short period of time or other LLM tells).
The real problem is that AI doesnt make any money. In fact, AI companies and Buisiness units hemmorage cash. When AI is eventually priced to the market cost the use-case for this all collapses.
OpenClaw Peter is using codex to analyze/de-duplicate PRs, extract good ideas from them and then re-implement them.
> I spun up 50 codex in parallel, let them analyze the PR and generate a JSON report with various signals, comparing with vision, intent (much higher signal than any of the text), risk and various other signals. Then I can ingest all reports into one session and run AI queries/de-dupe/auto-close/merge as needed on it.
Some people bitch, others are real engineers solving novel problems.
Often I see youtube videos that sells an overwhelmingly negative take on AI, like "OpenAI" fails 93% of Jobs or "AI is destroying the world" and other weirdly outlandish titles that is clearly aimed at clickbait.
Watching these content I often get confused because it never seems to highlight the actual real world progress and use that LLMs in particular gets for coding.
Much of what was "vibe coding" is becoming just coding now. This means for open source, we are no longer relying on companies that create "opencore" products that nerf/neglect the public version so they can sell their cloud product. We don't have to worry about a maintainer going AWOL on some Clojure or Elixir library and fret about hiring someone who has "20 years of experience". We don't need to pay for a lot of expensive enterprise SaaS tools that charge six digits when we can simply use LLM to internalize existing packages and even create our own.
Those that have been using coding agents for the past 6 month know how much progress there have been and the sheer pace of it to know that we are about to turn the corner, especially as new forms of computing are in the pipelines that will scale even faster without incurring more energy, moving away from text token gen to something else that humans can't read etc.
While it's important to watch different takes, I think someone who consumes only Youtube and these videos that the algorithm is designed to push is going to be shocked and left behind because by the time these videos are produced, things have already progressed or in state of change. All in all, these videos should be treated like ephemeral commentary that ultimately loses their relevance due to the sheer speed of how things are changing.
15 comments
[ 6.0 ms ] story [ 38.6 ms ] threadAuthor sounds like a relatively well off white dude in the 1950s.. 60s, 70s, 80s, 90s...
I get it, everything is being massively disrupted right now. I'm not trying to say ai is good or that bad, but the authors argument is weak.
It's just not clear to me who, or what, will do it.
> I spun up 50 codex in parallel, let them analyze the PR and generate a JSON report with various signals, comparing with vision, intent (much higher signal than any of the text), risk and various other signals. Then I can ingest all reports into one session and run AI queries/de-dupe/auto-close/merge as needed on it.
Some people bitch, others are real engineers solving novel problems.
https://x.com/steipete/status/2025591780595429385?s=20
Watching these content I often get confused because it never seems to highlight the actual real world progress and use that LLMs in particular gets for coding.
Much of what was "vibe coding" is becoming just coding now. This means for open source, we are no longer relying on companies that create "opencore" products that nerf/neglect the public version so they can sell their cloud product. We don't have to worry about a maintainer going AWOL on some Clojure or Elixir library and fret about hiring someone who has "20 years of experience". We don't need to pay for a lot of expensive enterprise SaaS tools that charge six digits when we can simply use LLM to internalize existing packages and even create our own.
Those that have been using coding agents for the past 6 month know how much progress there have been and the sheer pace of it to know that we are about to turn the corner, especially as new forms of computing are in the pipelines that will scale even faster without incurring more energy, moving away from text token gen to something else that humans can't read etc.
While it's important to watch different takes, I think someone who consumes only Youtube and these videos that the algorithm is designed to push is going to be shocked and left behind because by the time these videos are produced, things have already progressed or in state of change. All in all, these videos should be treated like ephemeral commentary that ultimately loses their relevance due to the sheer speed of how things are changing.