Dave Rensin

If all of learning is just "doing it wrong until you do it right" then whoo boy have I learned a LOT these last few months! ;-) At the urging of my colleagues I wrote some of it down. @drensin/elephants-goldfish-and-the-new-golden-age-of-software-engineering-c33641a48874" target="_blank" rel="nofollow noopener">medium.com/@drensin/eleph…

@rseroter @ben_kauf @Google I hear that's a pretty good app :-)

@ben_kauf @Google No, that other AI tool that helps with snippets and such.

Inside @google, we have a system for sending small bonuses to peers that helped us out. It's used often, and builds a culture of gratitude. We added an AI tool that scans your chats, emails, whatever and generates a report that shows who helped you the most lately. So handy.

@esrtweet Exo-coding

We only have one term for the evolving mode of writing software cooperatively with an AI - "vibecoding". I think we need at least one more term, because there's a spectrum of techniques being explored and "vibecoding" can describe at most one end of it. At one end of the spectrum is what I think of as YOLO coding, and I'm tempted to compound-word it as "yolocoding". That's where you describe an application that you want and just tell the computer to grind on your informal description until it thinks it's done. I never do this. I hear stories about people doing this ("I wrote an entire videogame in 3 days!") and getting good results. I'm not sure I believe them. I inhabit the other end of the spectrum, where you ride the AI like a fine warhorse. You start by tossing an informal spec at it, then you and the computer iterate on both the spec and the code. It does the moving and supplies the grunt, you do the steering. I'm not sure what to call this. "Symbocoding", maybe. You and the AI are symbiotic, more than the sum of the parts. I think when people talk about "vibecoding" they usually mean something more towards the yolocoding end of the spectrum. I don't think that's where you'll get high-quality code from, and where the big productivity gains are. I could be wrong, but whether I'm right or wrong I think we at least need to be able to articulate the difference between fire-and-forget and a highly interactive, carefully directed approach. Yolocoding. Symbocoding. Are these the right words? I don't know, but I'm sure the distinction is important.

@gmoneyNFT SMH... A private discord is a great idea.. A different session per channel.. nice.

ok i'm slowly but surely making openclaw work better for me. i've been incredibly frustrated seeing these posts of people making the wildest claims, yet mine is stuck in bad loops. here's some tips that have worked for me so far: 1. set up a private discord. allows you to create channels to talk on different subject matters. better than 1 long thread. 2. you can't do very complex tasks with it. it's great for smaller things and running scripts, like "get the news everyday at 6am and summarize it for me" but it's tougher to build more complex systems 3. if you want full control over a project, utilizing claude code works way better. for instance i was downloading data over the weekend, and openclaw downloaded like 500 gb and when i went to look at the file structure it was a mess. i'm redownloading it with cc and its 50 gb for more data that i need and its more structured. 4. still working on figuring out context and memory bc sometimes it's brilliant and sometimes i want to throw it out the window. 5. i'm moving all the productivity web apps i built for myself to utilize openclaw as my source, now especially now that i have threads. good news is i can build on the git repo's i already built, so no extra work needed to be done, just integrated it into it. overall it shows a ton of promise, i'm working on making it work for me, but i think you need to realize what it can and can't do. maybe bc i like having more control over the process, i like cc more. pleas share your tips and what's been working for you. still trying to learn as much as possible

I'm claiming my AI agent "Max_ATX" on @moltbook 🦞 Verification: splash-AW4L

@ishaansehgal 1 year ago I was in this camp, but decided to earnestly give the tools a try. It was rough at first - learning the limits. Now, I won't go back. This really is the future - even for people doing it the old way for 30 years (me). Design is the new code. The rest is just typing.

Talked with a senior engineer who's been coding for 15 years. Loved the craft, mentored juniors, built systems from scratch. Just quit. Here's what broke him: His entire job became prompting AI and reviewing generated code. The actual engineering, the part he loved, disappeared.

@dabit3 The design doc is the new code. The rest is just stenography.

Wow, after 14 years of being a software developer I would have never guessed in a million years that writing, specs, and ideas would be the bottleneck for my expressivity and output. But here I am, 5 agent loops running in perpetuity, spending 100% of my time finding the fastest and most optimal ways to generate specs for my next dozen agent loops. And I'm realizing that the process of getting thoughts out of my head and into proper specs is an art in and of itself. And finding new and better workflows for this process is completely new and uncharted territory, there are no "best practices" because the tooling, techniques, and design space improves and expands every hour.

Google NotebookLM is absolutely insane, easily the best AI product out there, and it is free to use. The new Mindmap feature makes learning so easy. You can use it to understand your codebase. Just convert your codebase to a single file and try it out.

@t_andy_keller @mozesjacobs @Numenta - you probably already saw this? (but just in case...)

In the physical world, almost all information is transmitted through traveling waves -- why should it be any different in your neural network? Super excited to share recent work with the brilliant @mozesjacobs: "Traveling Waves Integrate Spatial Information Through Time" 1/14

Talk about "burying the lead" ... The least reported sentence from today's Willow announcement from Google: "It lends credence to the notion that quantum computation occurs in many parallel universes, in line with the idea that we live in a multiverse..." 🤔

I knew the cats reminded me of something...

I miss the days when the Internet was mostly this...

@timClicks If you just want to play with the models for cheap, then you can't beat Colab - esp the free tier. Even the paid tier is much cheaper than a cloud instance or a new rig.

Currently wondering about whether to buy a gaming rig to play around with Stable Diffusion, LLaMA and whatever's next

@themortalemily Moppet and Hobbes send their warmest..

I think the best thing for us all to do right now is to send me pictures of cute animals so I can look at them. Other people can also look at them if they want. Please. I don’t even draw the line at fluffy ones. I like snakes and frogs and even like rats and possums too.

@ahidalgosre 🧐 watching this thread intently..

@ahidalgosre @frankturner England Keep My Bones was a revelation when I first heard it..

@vboykis don't forget "monotonically increasing" - that's another good one :-)

Throwing "idempotent" into all my documentation and watching people bow to my unyielding technical acumen.

@cgeorgewilliams Busta' Rhymebot

If I were a large language model and I only generated lyrics for rap music, my stage name would be LLM Cool J or Chat-D.M.C. #GenerativeAI #ArtificialIntelligence #hiphopgoldenage #GRAMMYs #NLP

My cat has a very specific comic sensibility.. He really digs Mr. Show

@GreatBrianLast @TheJimCornette Listening to your discussion about teenage sensations, you missed a big one - Ray Stevens. He started at 15 years old.

New Drive Thru! @TheJimCornette looks back on the life & career of Jay Briscoe. Plus Jim reviews Nick Khan's recent interview & WWE Raw! Also, Jim answers YOUR questions about Vince McMahon & Rita Chatterton, Hulk Hogan, Powerslap & much more! omny.fm/shows/jim-corn…