iamwil
@interjectedfuture
Tech Zine Issue 1: LLM System Eval Local-first/Reactive Programming ⁙ LLM system evals ⁙ Startup lessons ⁙ Game design quips. Longform: Podcast:
Not picking on this particular person. But I’m a little dumbstruck by this current discourse where people are all pikachu-face about how others weren’t lying, but the reason they were getting subpar results from AI was because it was a skill issue all along. x.com/karthiktade...
Out of anyone I track, @_chenglou really leans into the affordances of LLMs and then asks what we can do with this newfound affordance. I've adopted generating Quint proofs for complex state management. What other guardrails can we enable that was previous onerous? x.com/_chenglou/s...
For a long time, I also did not understand what value was and its relation to money flow. Now that I have a better understanding, I think this quip also applies to open source advocates and decentralization evangelists, not just mathematicians and students. x.com/soleio/stat...
I think @acadia.engineering is pointing at something, and HN's looking at the finger. I don't think this is really about Elm at all, but the preparation for Acadia.
This is an advantage of letting AI slurp up your work. Your defaults become in distribution and you don’t have to fight as hard for LLMs to “do it your way”. It’s the ultimate win in technical holy wars. x.com/mitchellh/s...
For example, how to find the right format for the spec? My answer: you don't. Let the agent interview you and decide the format for itself, because ultimately, it needs to follow the instructions and not you. bsky.app/profile/dnu...
Prep is where I started on my agentic coding journey. It's the work before the work to give the agent the right context. Do you have any questions or struggles I can try to address in the planned blog post? x.com/iamwil/stat...
Given how ahistorical tech is, I think it’s great when those who were in-the-room-where-it-happened drop in to set the record straight! bsky.app/profile/h.o...
I've also wondered as either an intentional or unintentional attack vector. I'm glad someone's asking the question, as it's a good start. x.com/kanjun/stat...
If you currently agentic code by iterating back and forth with the agent to get a result, then this is an understandable interpretation of loop engineering. But if you already do this, then you're ready for cybernetic loop engineering.
My fun fact for this is that we're closer in time to the T-rex than the T-rex was to the Stegosaurus. By a large margin of ~20 mil years. Another: Anne Frank, Martin Luther King Jr, and Barbara Walters are all born in the same year (1929), even though we associate them with different eras.
I keep referring to 3 tribes. josephg.com/blog/3-tribes/ This is the maker tribe disparaging the poet tribe. It certainly is true that the poet is prone to navel gazing. But it's also true that the poet tribe is a rich source of new materials the maker tribe then gets to use. x.com/awesomeklin...
Lots of ink spilled on "Loop engineering", but no one seems to connect it to its much older name: Cybernetics. Maybe by the year's end, we'll recognize that it was all Control Theory engineering all along.
That said, the thread's overall point is correct. The idea of printing docs to communicate in an org is vestigial now. It make more sense today if docs could be just as easily read by machines. Hence, markdown today with LLM-based agents. x.com/palashshah/...
Lots of people worked hard so that later on, we can have the luxury of generational amnesia. Docs definitely weren't easy to print before PDF. One example: you didn't have a font on your machine, it just printed garbage. "It worked when I printed it" was a real thing. x.com/palashshah/...
Would a company be able to properly discern? I suppose so. Interviews and resumes are a proxy for work. So if AI amplifies work and it's demonstrable that an employee outperforms their peers, then it raises the SNR. x.com/howard/stat...
So it is with skill packs. I think this probably transfers to the burgeoning marketplace of agent skills that people trade and adopt with each other. Any contemporary anthropologists should be interested in tracking this. Tell your anthro spouses, tell your anthro friends.
There's also Ink and Switch's CrossCut. www.inkandswitch.com/crosscut/ x.com/iamwil/stat...
All in response to this one. To which I say, my experience is such that I've been cringing at reading my coworker's code for a while now. And it's also within the realm of possibility you haven't developed taste for code yet, regardless of how long you've been writing it. x.com/0interestra...
I present, "Comedian" by Maurizio Cattelan. How do modern art Conceptualists fit into this worldview that you need skill to be an artist? x.com/NJSimmondsb...
More apt comparison is the camera. Artists definitely had to wrestle with that one, before accepting it into the pantheon of creativity. x.com/ABAOProduct...
> coding is a means to an end, invented because it was a tool. I see you've never met any category theorists on Twitter yet. x.com/Baxate/stat...
Today, I make a rare reaction video on AI and artist takes on Twitter. x.com/Issybeatz_/...
I strongly think there's a gap in the market for "dumb devices", or at the very least, local-first devices that you know a) survives network/server outages and b) is on your side(no spying). I think there's a unit econ/business model issue though. x.com/internetofs...
Historic development of technology would be fun cases to do product positioning and segmentation. Would you have identified the gap a bus would have filled? Would a historical LLM you're working with?
Oddly, there are no tech podcasts like there are science podcasts, like Radiolab, Veritasium, or Ologies. Most technology podcasts are too facile, or too gossipy, or for insiders. There are no tech podcasts that explain the modern world to a reasonably curious person. x.com/iamwil/stat...