Reposted by sawyerHannah Shelley, MLIS (Maggots, Lycanthropes & Ice Skeletons) @hannahshelley.site · 02/10/2026Please stop anthropomorphizing computers. They cannot "run" a program, because they don't have legs 392065287
sawyer @sawyer-p.me · 02/10/2026At work am increasingly needing to think in terms of billions and trillions, which is uh 😳 010
sawyer @sawyer-p.me · 28/09/2026I think I have had 4 interviews in the past year where the company is up front in the first screening call that they are 6 days a week or 7(!). Bay Area. I have had even more interviews where screener asks “Tell me about a time that you worked long hours for an extended period of time” 🫠 001
Reposted by sawyerIsaiah Bishop @isaiahbishop.bsky.social · 17/09/2026What jev promises is an essentially classifier for any problem and that rules 3865
Reposted by sawyerMatt Henderson @matthen.com · 11/09/2026A cellular automaton from Minecraft crafting recipes 522633
sawyer @sawyer-p.me · 08/09/2026Finally, but not going nearly far enough. I want the right to deny the use of both data structures AND algorithms. Rampant use of linked lists by these sites is frankly disgusting. www.theguardian.com/australia-ne...theguardian.comAustralia to force social media platforms to allow users to opt out of algorithmsDraft legislation enforces digital duty of care including limiting children’s exposure to misogynistic and eating disorder content 110
sawyer @sawyer-p.me · 08/09/2026I have certainly bricked my Linux machines many a time back in the day while tinkering. These days using NixOS I have full safety for pretty much any modification I want to make. 011
sawyer @sawyer-p.me · 08/09/2026Dotfiles are a configuration surface. One thing you might want your distro to offer is the ability to roll back to a previous incarnation of your config. Many distros offer something like this. NixOS being maybe the most robust incarnation of the idea. 130
sawyer @sawyer-p.me · 08/09/2026Comparing brains and LLMs in terms of information processing is an actually tractable problem. I don’t know much in this area but I imagine we can at least make some useful insights. It’s measurable. An entities ability to experience qualia, at least as far as I know, we pretty much have no angle on 000
sawyer @sawyer-p.me · 08/09/2026I know there are theories of consciousness that try to take advantage of information theory. I don’t find these particularly useful for understanding where qualia comes from. I think the honest answer is we are very far from figuring out the hard problem of consciousness, the origin of qualia. 130
sawyer @sawyer-p.me · 08/09/2026LLMs as information processing systems are quickly approaching being much better than humans under nearly all measurable axes. Which comes with a sense of despair. But lumping in consciousness, the capacity to experience qualia, as being associated with processing ability is dubious. 100
sawyer @sawyer-p.me · 08/09/2026Oh my god this is an all-timer Jerma stream. …im crying laughing youtu.be/xuswv1O9F4c?... 110
sawyer @sawyer-p.me · 08/09/2026Not to mention it selling itself as an “agentic” distro despite having, i don’t know, seemingly no features that you want if AI is configuring your system. There’s enormous prior art here. 160
sawyer @sawyer-p.me · 08/09/2026I would say it’s definitely made many parts of programming easier, not just generating code. System design and code review have genuinely become easier now. AI systems are actually incredible as a partner to help you understand new code and design new systems. 010
sawyer @sawyer-p.me · 08/09/2026Emacs with the ECA package is pretty amazing. Have not used emacs in a while and forgot just how well designed it is. How has every single piece of software been so CONSISTENTLY wrong about how to make a productive text based computer interface. 010
sawyer @sawyer-p.me · 07/09/2026People be talking about making new terminal multiplexers when all we really need is a super performant emacs you can't accidentally lock up by writing stupid elisp. 010
sawyer @sawyer-p.me · 07/09/2026(Interstellar theme no time for caution blasting as the locomotive accelerates to 15 mph) ITS NOT POSSIBLE No, it’s necessary 011
Reposted by sawyerjoe is afraid @crimson2877.bsky.social · 07/09/2026My article is up! retroobserver.com/theres-still...retroobserver.comThere's Still Nothing Like Romancing SaGa After 30 YearsRomancing SaGa on the Super Famicom is an unfairly forgotten gem, unique from its remake and loaded with intrigue 2168
sawyer @sawyer-p.me · 07/09/2026People on sailboats will be hootin and hollering "holy shit we're going FAST" and it's like 7 mph 220
sawyer @sawyer-p.me · 05/09/2026You can do that way more effectively if you’ve modularized and decoupled your system. You can also do that way more effectively if you hand wrote at least all the types and function stubs! 000
sawyer @sawyer-p.me · 05/09/2026Telling an LLM “go solve this constrained problem with these tools this is how you know you’ve done it right” is actually a pretty great experience these days. 100
sawyer @sawyer-p.me · 05/09/2026Building on this, programming the actual system architecture by hand I kind of feel is just as (if not way more) efficient than looping on design docs with LLMs. Really really good architecture lets you confidently delegate work to LLMs. The fundamental patterns and system design are so important. 110
sawyer @sawyer-p.me · 05/09/2026Some weird and interesting things from this week: 1. Wrote some code by hand today, did not feel weird or uncomfortable at all. Expected some atrophy, but noticed none. 2. Shipping a major new product at work heavily built with AI. Some issues but far more on the rails than expected. 010
sawyer @sawyer-p.me · 02/09/2026Designing architecture and code patterns is distinct from actually implementing them. Iterating on these designs in a loop with an LLM is very effective. Also noting that this is quite literally an ML problem. 100
sawyer @sawyer-p.me · 02/09/2026Learning how to constrain them is not magic, it’s just an ML problem. It’s not about prompting either. It’s about making a system with excellent deterministic quality checks, paths for human feedback, and excellent code design and architecture to properly constrain the agents. 100
sawyer @sawyer-p.me · 02/09/2026I think it’s important to think about when and how humans are pulled in to do meaningful design and architecture work. LLMs cannot produce well engineered things without lots of guardrails and guidance. But they can work, and if constrained correctly, work astonishingly well. 100
sawyer @sawyer-p.me · 02/09/2026Yeah that’s precisely why it’s difficult. The models are very smart but have no memory or concept of your system if you directly invoke them with Claude code/codex. A non trivial engineering effort is needed to solve the problems you’re calling out here. 100
sawyer @sawyer-p.me · 02/09/2026A year ago I definitely agreed, but it’s genuinely possible now to build software both better and faster with the use of LLMs. It’s very difficult to get that Pareto improvement, but possible. You go from engineering the code to engineering the machine that can produce and validate the code. 110
Reposted by sawyertweety fish oh wow like a chimera or something spoooky @sifu.tweety.fish · 02/09/2026there is nothing whatsoever inherent to data centers or the kind of computation necessary for LLMs that means they have to be polluting; they're a great option for green energy for various reasons and can be closed loop with non-potable water etc. [ CUT TO: PRESENT MOMENT ] ah well nevertheless 410615
sawyer @sawyer-p.me · 02/09/2026Advertised as an “agentic” distro Is not versioned and immutable, no rollback mechanism at all lol 030
sawyer @sawyer-p.me · 02/09/2026Why is it that corporate sponsors are lining up behind a Linux rice. 110
sawyer @sawyer-p.me · 02/09/2026Damn literally just set up my entire nix system around 1Password… it’s honestly crazy that a fascist Linux rice has corporate sponsors. Not even really a distro! 020
sawyer @sawyer-p.me · 31/08/2026Yeah for sure. Want to explore this some more. Bigger models love to bike shed on super edge case issues which really don’t matter early on in design. I can definitely see how a dumber robo rubber duck is more valuable when you’re just thinking through something. Just need a little entropy. 120
sawyer @sawyer-p.me · 31/08/2026That’s interesting. I find myself doing pretty intensive long design sessions before being able to see things clearly enough to break things into a straightforward task graph. I do find sol being more useful/mature than alternatives during this phase? Luna can crush individual tasks. 110
sawyer @sawyer-p.me · 30/08/2026All I watch is The Rest is History episodes about some 2k year old beef between the Persians and the Greeks 010
Reposted by sawyerCosmilk @cosmilkmoosic.com · 28/08/2026My new track "From Ash" with Hollow Mage is out now as a part of Unbeyond Anthology Three! Would love it if you gave it a listen! @unbeyondanthologies.ca soundcloud.com/unbeyond-ant...soundcloud.comCosmilk & Hollow Mage - From Ash“Unbeyond Anthology Three” is OUT NOW on Unbeyond Anthologies! Buy / Stream ↣ https://hypeddit.com/uba075 @cosmilkmoosic @hollowmage ---------------------------------------------------------------- 1116
sawyer @sawyer-p.me · 29/08/2026Lord please forgive this goofiness, bring shame upon my laugh and smiles. 020
sawyer @sawyer-p.me · 28/08/2026Reading the classics > We should recognize that already now programming is much more an intellectual challenge: the art of programming is the art of organizing complexity, of mastering multitude and avoiding its bastard chaos as effectively as possible. Notes on Structured Programming, 1969 000
Reposted by sawyerBryan Cantrill @bcantrill.bsky.social · 06/12/2025Your intellectual fly is open bcantrill.dtrace.org/2025/12/05/y...bcantrill.dtrace.orgYour intellectual fly is open | The Observation Deck 815834
Reposted by sawyerAbel @abeliansoup.bsky.social · 23/08/2026have you ever seen a visual shepherd tone anon? courtesy of @bleuje.com 1431063
Reposted by sawyerLouie Zong @everydaylouie.bsky.social · 22/08/2026nirvanna the band the show the fanart 113942741
sawyer @sawyer-p.me · 21/08/2026I live in Plato’s cave but instead of seeing shadows all I see are slack messages 130
sawyer @sawyer-p.me · 20/08/2026Razor love is one of his beststatic.klipy.comSmall Bart Simpson Rotating LoopALT: Small Bart Simpson Rotating Loop 010
