conputer dipshit
LightNews LightNews
banner
davidcrespo.bsky.social
conputer dipshit
@davidcrespo.bsky.social
1.4K followers 510 following 4.5K posts
web dev + hot dad. enjoy charts, unions, conputer games, philosophy. chicago crespo.business
Posts Replies Media Videos
davidcrespo.bsky.social
I've been using them pretty much daily in a non-agentic chat CLI for 2 1/2 years. from a 2025 POV, I would describe GPT-4 as basically a piece of shit, and even then it was useful every day!

here is a useful chat I had from that time about web dev
gist.github.com/david-crespo...
github commits from april 2023 use gpt-4. big spender david-crespo committed on Apr 19, 2023 Commits on Apr 18, 2023 rudimentary history concept david-crespo committed on Apr 18, 2023 jsonMd david-crespo committed on Apr 18, 2023 chatgpt script works to my liking david-crespo committed on Apr 18, 2023 initial commit david-crespo committed on Apr 18, 2023
November 10, 2025 at 5:41 PM
davidcrespo.bsky.social
or not! what is one supposed to do with this? I've been using claude code most days for like 6 months. every day a new coworker (I work with the best engineers you've ever seen in your life) is like "oh neat LLMs work now"
‪your #3 source for absurdist true crime 🔨‬ ‪@davidgerard.co.uk‬ · 7h The key weakness in AI agents is that they're a lie. They don't work. They just don't fuckin' work. You can't set a hallucination engine to work doing tasks. It's pants on head stupid. The hype pretends this isn't the case and hypothesises a fabulous future where they work *at all*. This is a lie. ‪your #3 source for absurdist true crime 🔨‬ ‪@davidgerard.co.uk‬ · 7h A useful model for "AI agents" is that they're the current excuse meme for AI. They're not a thing that works at all, now or in the fabulous future. But they're *such* good material for hypecrafting.
November 10, 2025 at 5:34 PM
davidcrespo.bsky.social
this is good, but what will really cook a libertarian's brain is that even under capitalism, markets are a creation of the state
www.vox.com/politics/467...
Question: You are a proponent of “market socialism.” To many, that may sound like a contradiction in terms. Historically, socialism has been associated with the ambition to cease allocating resources through markets and begin doing so through central planning. So why do you think that even a socialist utopia would want to preserve the market economy? Answer: The market in and of itself isn’t capitalism. Capitalism is a system where firms hire wage labor to produce for profit, extracting surplus value and reinvesting it to expand their capital. In my vision, a socialist society would retain markets as tools for coordination and consumer choice. But the firms operating in these markets would be run by worker cooperatives that are funded by public banks. The reason why I’ve come to embrace markets is just by looking at the actual practice of central planning under socialism. In any system, society still has to decide what to produce, how much, and for whom. Central planning attempted to answer these questions administratively. But planners lacked the information necessary to rationally allocate goods. And they also faced severe incentive and agency problems — managers had little reason to report results honestly or innovate when the survival of their firms didn’t depend on efficiency.
November 9, 2025 at 5:34 AM
davidcrespo.bsky.social
MAINFRAME for ZOOMERS
November 9, 2025 at 4:59 AM
davidcrespo.bsky.social
going with the first thing that came to mind
homes.luddy.indiana.edu/ccshan/for/f...
BUTT for POOPING, SILLY in the style of a zohran sign
November 9, 2025 at 4:57 AM
davidcrespo.bsky.social
he might want to give that AI thing a try
terminal output $ ai why might cumulative net income differ from change in net cash sonnet-4.5 | 10.49s | $0.00592 | Tokens: 229 -> 34 Cumulative net income and change in net cash differ because net income is calculated on an **accrual basis** while cash flow tracks actual cash movements. Several key factors drive this divergence: **Non-cash expenses and income**: Depreciation, amortization, and stock-based compensation reduce net income without affecting cash. Conversely, non-cash gains (like asset revaluations) increase income without generating cash. **Timing mismatches**: Revenue is recognized when earned, not when cash is collected. A company may record a sale (increasing net income) while the customer pays 60 days later (delaying cash receipt). Similarly, expenses are recognized when incurred, not when paid—accrued liabilities reduce net income before cash actually leaves. **Working capital changes**: Increases in accounts receivable or inventory consume cash without impacting net income. Building inventory requires cash outflow but doesn't affect earnings until those goods are sold. Conversely, increases in accounts payable provide cash while not changing net income. **Capital expenditures and financing activities**: Purchasing equipment is a cash outflow that doesn't immediately hit net income—it's capitalized and depreciated over time. Similarly, debt repayments, dividend payments, and equity issuances all affect cash but bypass the income statement entirely. **One-time items**: Gains or losses from asset sales, restructuring charges, or legal settlements can significantly impact net income in a period while having different (or delayed) cash implications. This is why the statement of cash flows exists—to reconcile net income with actual cash generated or consumed during a period.
November 9, 2025 at 2:54 AM
davidcrespo.bsky.social
bonus points for funniest possible choice of shirt
suburban mom being arrested in pink shirt that says "TOXICALLY empathetic"
November 8, 2025 at 1:42 AM
davidcrespo.bsky.social
this is going to sound insane but I have long dreamed of a sort of science of puzzle state space that would both inform puzzle design but also enable the design of new puzzles that are played by jumping between levels of abstraction
worrydream.com/LadderOfAbst...
Three Steps Up The higher we climb, the broader our view of the system, and the more high-level patterns we can notice. Here is a third-level abstraction — a representation of the system for all time, all turning rates, and all bend angles. The color represents the time to completion, from red (fast) to black (mid-range) to white (slow). We're pretty far up in the clouds now — this representation bears no resemblance to the motion of a car, and omits most lower-level details. But as you should expect, those details are available by interactively stepping down. Mouse over the chart to step down two levels, and see the trajectory for that particular turning rate and bend angle. Look for pixels whose color is conspicuously different from their neighbors, and try to figure out why they stand out. Mouse over the y-axis to step down just one level, and see trajectories for that particular turning rate. Mouse over the x-axis to step down one level, and see trajectories for that particular bend angle. Each abstraction from earlier can be seen as a slice of this third-level abstraction.
November 7, 2025 at 4:54 PM
davidcrespo.bsky.social
there's just something about people born between 1988 and 1991, isn't there!

(the graph is probably misleading, I bet most of the improvement was in the first half of the 80s)
graph of lead concentration in the blood of children in the US from 1978 to 2016 showing a steep decline from 1978 to 1990
November 6, 2025 at 7:33 PM
davidcrespo.bsky.social
get in losers we're doing the nordic welfare state
www.nytimes.com/2025/11/05/n...
Others consulted by Mr. Mamdani and his aides include Marshall Ganz, a professor at Harvard University and an expert in community organizing; Matt Bruenig, a labor lawyer who founded a left-leaning think tank; and Lilliam Barrios-Paoli, a deputy mayor for health and human services under Mr. de Blasio.
November 6, 2025 at 7:09 PM
davidcrespo.bsky.social
as I say every week it feels like, the danger is not that they will be inaccurate. it is that they will be accurate. and they will be
www.404media.co/ai-is-superc...
“I see something like this, and it’s presented as very value neutral, like, ‘Here’s something that is going to make life easier for you because you have all these books you need to review,’” Jaime Taylor, discovery & resource management systems coordinator for the W.E.B. Du Bois Library at the University of Massachusetts told me in a phone call. “And I look at this and immediately I am seeing a tool that’s going to be used for censorship because this large language model is ingesting all the titles you have, evaluating them somehow, and then it might spit out an inaccurate evaluation. Or it might spit out an accurate evaluation and then a strapped-for-time librarian or teacher will take whatever it spits out and weed their collections based on it. It’s going to be used to remove books from collections that are about queerness or sexuality or race or history. But institutions are going to buy this product because they have a mandate from state legislatures to do this, or maybe they want to do this, right?”
November 6, 2025 at 5:59 PM
davidcrespo.bsky.social
one of my professors in college, Arnold Davidson, quoted an unpublished line from Foucault that I think about all the time: "the art of living is to kill psychology"
www.cambridge.org/core/books/a...
for cutting" and for "shattering" (EFR, 88). Foucault does not aim to establish new knowledges but to cut and shatter what we currently consider knowledge. The psychological sciences, and first and foremost psychiatry, were among the things that Foucault cut up, shattered, and attacked. In conversation, he once said that "The art of living is to kill psychology... If one cannot manage to do that in life, it isn't worth living" (FDEza, 1075, my translation). Foucault's final "ethical" works on the aesthetics of the self can thus be read as attempts to kill psychology or, more modestly, to find and describe alternatives to the psychologized self. These works show that
November 6, 2025 at 2:49 PM
davidcrespo.bsky.social
"We construct (in red) an intelligence test (horizontal), that is 100% correlated with negative performance (when IQ is, say, below 100) and 0% with upside, positive performance. We progressively add noise (with a 0 mean) and see correlation (on top) drop but shift to both sides."
Fig 1: The graph that summarizes the first flaw (assuming thin tailed situations), showing that “correlation” is meaningless in the absence of symmetry. We construct (in red) an intelligence test (horizontal), that is 100% correlated with negative performance (when IQ is, say, below 100) and 0% with upside, positive performance. We progressively add noise (with a 0 mean) and see correlation (on top) drop but shift to both sides. Performance is on the vertical axis. The problem gets worse with the “g” intelligence based on principal components. By comparison we show (graph below) the distribution of IQ and SAT scores. Most “correlations” entailing IQ suffer the same pathology. Note: this is in spite of the fact that IQ tests overlap with the SAT! (To echo Haldane, one ounce of rigorous algebra is worth more than a century of verbalistic statisticopsycholophastering).
November 6, 2025 at 4:33 AM
davidcrespo.bsky.social
I think about this miami sales grifter girl all the time. I do love to yap
www.tiktok.com/@shelbysapps...
if you're a coder and you love to yap you need to get your butt into claude code with macwhisper
November 5, 2025 at 10:48 PM
davidcrespo.bsky.social
was curious if there was anything redeeming in the context, but the bit immediately preceding suggests it's impossible to know how Mamdani would have done if it was just him against a Republican. are you insane?
The New York mayoral race doesn’t clarify things, either. In many respects, Mr. Mamdani’s election is extraordinary, and the enormous youth turnout for his candidacy is extraordinary as well. When the data is final, it will probably show that the share of voters 18 to 29 approached or perhaps even exceeded its share from the last presidential election. This is unheard-of. It’s the kind of youth surge that the Bernie Sanders campaign promised, but never delivered. But with Andrew Cuomo — a Democrat — running as an independent, the race did not ultimately test Mr. Mamdani’s strength in a one-on-one contest against a Republican. The Democratic vote was split, and it’s impossible to know how many of Mr. Cuomo’s traditionally Democratic supporters would have voted for the Republican. It’s possible most Cuomo voters would have come over to his side, but, again, we won’t know.
November 5, 2025 at 7:35 PM
davidcrespo.bsky.social
fun shot of the initial test, where I told it to open a file, type in some invalid code, and confirm that the language server error pops up
• tmux is actually well-suited for this - tmux capture-pane lets me for observing LSP errors. Let me try: see what's rendered on screen, which is crucial • Bash(tmux new-session -d -s helix-test) L (No content) • Bash(tmux send-keys -t helix-test 'hx helix/init.scm' C-m) L (No content) • Bash(sleep 0.5 && tmux capture-pane -t helix-test -p) L (require "helix/configuration.scm") (require "helix/keymaps.scm") 3 (require (prefix-in helix. "helix/commands.scm")) ... +21 | lines (ctrlto to expand) • Excellent! Helix is running. Now let me enter insert mode and type something: • Bash(tmux send-keys -t helix-test 'i') L (No content) • Bash(sleep 0.2 && tmux capture-pane -t helix-test -p) L 1 (require "helix/configuration.scm") 2 3 ... +20 #%exp-picker lines (ctrlto to expand) variable | • Perfect! Insert mode is active (see "INS" in status line) and autocomplete appeared. Now let me type some invalid code: • Bash(tmux send-keys -t helix-test Escape 'G' 'o' '(abc)' Escape) L (No content) • Bash(sleep 0.5 && tmux capture-pane -t helix-test -p) 1 Go(abc) (require "helix/configuration.scm") Lfree identifier: abc -free identifier: Go ... +20 | lines (ctrlto to expand) • Excellent! This works. I can see: 1. LSP errors are visible: Line 1 shows • marker with inline diagnostics: "free identifier: abc" and "free identifier: Go" 2. Status line shows error count: "• 2 W • 2" indicates 2 warnings/errors 3. Real-time feedback: The LSP analyzed the invalid code and displayed diagnostics The approach works for editor interaction. Let me clean up and close: • Bash (tmux send-keys -t helix-test ':q!' C-m) L (No content) • Bash(tmux kill-session -t helix-test) L (No content) • Answer: Yes, this approach works. Using tmux with send-keys and capture-pane successfully allows interaction with Helix including: - Entering/exiting insert mode - Typing text that triggers LSP analysis - Observing error diagnostics rendered in the editor UI (inline errors, status line indicators) This means validatio…
November 5, 2025 at 7:06 PM
davidcrespo.bsky.social
credit to @thdxr.com for the idea
x.com/thdxr/status...
November 5, 2025 at 6:44 PM
davidcrespo.bsky.social
I figured out how to get Claude Code to interact with the Helix editor directly through tmux, which means it can iterate on config changes (keymaps, language servers, formatters, etc) and test *directly* that they work. this is wild

rough early prompt here:
gist.github.com/david-crespo...
November 5, 2025 at 6:38 PM
davidcrespo.bsky.social
take exit polls with a grain of salt but this is pretty good
www.cnn.com/election/202...
Gender by race 4,744 total respondents White men 23% White women 27% Black men 7% Black women 10% Latino men 7% Latina women 10% All other voters 16% Mamdani 42% 47% 62% 54% 55% 51% 60% Sliwa 9% 8% 5% 3% 11% 7% 7% Adams 0% 0% 1% 1% 0% 0% 0% Cuomo 49% 45% 31% 42% 33% 42% 32% Updated 1:01 p.m. ET, Nov. 5 Age 4,744 total respondents 18-29 11% 30-44 24% 45-64 33% 65 or older 31% Mamdani 78% 66% 43% 36% Sliwa 4% 5% 9% 8% Adams 0% 0% 0% 0% Cuomo 18% 28% 47% 55% Updated 1:01 p.m. ET, Nov. 5
November 5, 2025 at 6:26 PM
davidcrespo.bsky.social
lol
If you haven't played Q-UP before, here's what our game is all about: 🪙 You flip a coin 🪙 You can flip a coin with your friends 🪙 You can flip a coin on the internet 📈 Hundreds of skills, items, 8 unlockable characters, a full incremental RPG with a zachtronics-style PoE-like factory engine building system, a deep immersive story, etc. 🪙 If the coin lands on your side, you win the coinflip! 🪙 Experience the thrills and chills of a 50% chance of something happening We can't wait to see you guys in the ARENA -james, frank, joon & martin
November 5, 2025 at 6:23 PM
davidcrespo.bsky.social
I *knew* I recognized millennial queen Lina Khan behind him. my god that rules
www.nytimes.com/live/2025/11...
Speaking at his first news conference as mayor-elect, Zohran Mamdani announced his transition team. Its leader: Elana Leopold, a de Blasio administration alumna. The co-chairs will be Lina Khan, the progressive former Federal Trade Commission chair, and Maria Torres-Springer, a former first deputy mayor under Eric Adams. The transition will be advised by Melanie Hartzog, a former deputy mayor for health and human services, and Grace Bonilla, the head of United Way of New York City.
November 5, 2025 at 5:54 PM
davidcrespo.bsky.social
seems like cuomo really did have a shot if they could have gotten sliwa to drop, though not all of his voters would have gone to cuomo. so it was not fully insane for him to do what he did. except everyone hates him even more now. a postponed comeback might have worked better
November 5, 2025 at 2:36 AM
davidcrespo.bsky.social
betting odds went straight to 99.6 on the initial results
polymarket.com/event/new-yo...
November 5, 2025 at 2:09 AM
davidcrespo.bsky.social
things are truly insane over there
Rep. Andy Ogles @RepOgles · 12h WAKE UP NEW YORK! embedded video of the second plane hitting the world trade center
November 4, 2025 at 3:16 PM
davidcrespo.bsky.social
fully half of my twitter algo feed is this kind of thing. it is nuts
November 3, 2025 at 9:20 PM