Posts on X before Nov 1, 2024, 7:14:22 AM UTC
@MichaelStolarzthey need to add a chatgpt search filter to the chat history ASAP, the sidebar is just spammed now with my quick searches and it's only been a few hours so far of use from me
@MichaelStolarzshopping is already a seamless experience with chatgpt search btw, don't you think they'll make bank on the conversions from chatgpt -> straight into the website to buy through affiliate/referral/marketing id links?
@MichaelStolarzone chatgpt search downside i've noticed (though it's more annoying than "wrong")
- it repeats itself on followup queries
- it still answers the new question and finds new sources again, so that's good
- seems like an easily solvable problem to de-duplicate
@MichaelStolarzi always wonder how much planning goes into cool little one-off halloween/etc holiday fun videos like this
like, was this always meant to be part of the functionality, or did brett think "this would be cool for halloween, MAKE IT HAPPEN ENGINEERS!"
and will it be useful again?
@adcock_brett
We did a Thing! ✋
@MichaelStolarzyeah... im cancelling my perplexity subscription, openai's search is game changing.
@MichaelStolarzLuce is going to be the largest positive long-term influence for catholicism from the vatican since... probably since constantine's conversion. it's going to take catholicism to (even higher) mainstream levels than ever previously imagined.
few.
@MichaelStolarzok, how about - get this - christianity?
@tsarnick
Bill Gates says AI will become so good at solving problems and creating attractive activities for humans to do that we need a new religion or philosophy to stay connected to one another
@MichaelStolarznew net worth goal: $1 decillion
@CultureCrave
Russia is trying to fine Google $20 decillion over YouTube bans • The fine surpasses the entire wealth and asset value on Earth • Google so far has ignored their demands
@MichaelStolarzinsane to me that @cursor_ai STILL doesn't have a commit message generator button like copilot does. is it patented or something?! seems like a no-brainer feature, and also seems like cursor deliberately isn't adding it in for some reason.
@MichaelStolarzprobably because python is the best performing language for LLM coding success
@ashtom
Even in this, Python is still growing faster than both JS and TS combined! So there’s cause for celebration from the Pythonistas 🐍🐍🐍 x.com/BenLesh/status…
@MichaelStolarzfor the past year or so, 99% of the time the answer to any question like "I'm having a hard time understanding ___, what should I do?" is: Ask a SotA LLM. The latest Sonnet 3.5 from Anthropic, 4o or o1 from OpenAI. They'll explain + create example problems for you. Every single… Less
for the past year or so, 99% of the time the answer to any question like "I'm having a hard time understanding ___, what should I do?" is:
Ask a SotA LLM. The latest Sonnet 3.5 from Anthropic, 4o or o1 from OpenAI.
They'll explain + create example problems for you. Every single time.
@MichaelStolarzif you use LLMs primarily for coding purposes, you're wasting your time looking at any other benchmark other than Aider's. let me explain and show you the best to track. first, a friendly reminder that chatbot arena scores are, for the most part, retarded, most likely skewed wi… Less
if you use LLMs primarily for coding purposes, you're wasting your time looking at any other benchmark other than Aider's.
let me explain and show you the best to track.
first, a friendly reminder that chatbot arena scores are, for the most part, retarded, most likely skewed with sophisticated bot systems to increase votes (and if not, then this is a missed opportunity for all the big labs, but i digress-)
most most importantly - llm leaderboards like lmsys/lmarena/chatbot arena are based on vibes and how cool the output looks (formatting, speech style) instead of "correct" results
the only benchmark i've found to be consistently reliable in how i experience the quality of today's SotA LLMs is the Aider code editing benchmark from @paulgauthier, as well as his second refactoring-based coding benchmark (which is way more demanding).
this is exactly the order of LLMs I would use if the best one were unreachable for some reason. No sonnet? o1-preview then. no o1-preview either? OK, gpt-4o is fine too (opus if the price was the same). and so on. the list is just... perfect.
@arena
Chatbot Arena Update🔥 @AnthropicAI latest Claude 3.5 Sonnet, has been extensively tested in Arena, securing an impressive #6 overall and #3 under style-control! With over 7K community votes, the new Sonnet is showing exceptional strength across various domains. Highlights: - H… Less
Chatbot Arena Update🔥 @AnthropicAI latest Claude 3.5 Sonnet, has been extensively tested in Arena, securing an impressive #6 overall and #3 under style-control! With over 7K community votes, the new Sonnet is showing exceptional strength across various domains. Highlights: - Hard Prompts: #4 (#1 with SC) - Coding: #2 (#1 with SC) - Math: #3 Congrats to @AnthropicAI on the impressive new release! More analysis below👇
@MichaelStolarznew reason to vote for trump: if kamala wins, geohot isn't coming back to america. GEOHOT. it's like if white zombie lost rob zombie.
@geohotarchive
America’s Future geohot.github.io//blog/jekyll/u…
@MichaelStolarzthis year was @rlgrime's last halloween album release (13/XIII); time to bring it back all the way to the beginning. let's listen to and rate all of them from the start. I'll rate each on a scale of 1-13 (since there are 13) 🧵 alright, so first up is... the first one. opening b… Less
this year was @rlgrime's last halloween album release (13/XIII); time to bring it back all the way to the beginning. let's listen to and rate all of them from the start.
I'll rate each on a scale of 1-13 (since there are 13) 🧵
alright, so first up is... the first one. opening by rl stine the goat himself. you read that right - RL STINE, the book writer, someone I grew up with during 7yo-14yo reading his work.
anyway - now at the time of writing this, I'm listening through Halloween I, then I'll listen through Halloween II, and make a new tweet in this thread with the updated list...
And so on, and so on, until all 13 are listened to and rated.
RATING SCALE
(1st place is best, 13th place is worst):
1st: Halloween I
2nd: NA
3rd: NA
4th: NA
5th: NA
6th: NA
7th: NA
8th: NA
9th: NA
10th: NA
11th: NA
12th: NA
13th: NA
@MichaelStolarzthis seems insane. i do 15 grams of coffee per 400 grams of water. am i the retarded one or?...
@MichaelStolarzwait a second, we're supposed to check the code? it just works for me every time, and when it doesn't, i tell the llm it doesn't work, and keep doing that until it works. @yacineMTB
@MichaelStolarzI pray this to be the case, and not the alternative nightmare scenario: we're so in our internet culture bubbles, we think this to be true
I hope Brian is right here!
@brian_armstrong
This seems like the first election where new media has fully flipped traditional media. Long form podcasts, X/social, prediction markets etc deciding this election. Also holding traditional media accountable. Happened gradually then suddenly.
@MichaelStolarzwoah - it would be interesting to see what an algorithm based on "verified engagement only" would look like. i'd love a toggle for that
@doganuraldesign
X should show verified engagements.
@MichaelStolarzanother way to bypass these safeguards is to use google instead
@newscientist
AI models have safeguards in place to prevent them creating dangerous or illegal output, but it turns out that writing backwards can be enough to trick them into revealing bomb-making instructions. newscientist.com/article/245083…
@MichaelStolarzsonnet 3.5, especially the new version of it, is GODLY at ui. the original sonnet was already really good, but this new update to it has great spatial awareness (or, css awareness I guess)
@MichaelStolarzAmen. t.co/nvLSD1dBx3
@MichaelStolarzhonestly might spend a week building an ai doc updater instead of actually updating these docs. who actually enjoys writing documentation? i want to keep building, no stop. documentation is my villain origin story at this point
@MichaelStolarztrying to learn rust again but my brain keeps going "skill issue" every time i fight the borrow checker
@MichaelStolarznew experiment: removing youtube as a bookmark. i find youtube becoming more of a distraction recently rather than a place where i go to learn something new; now that perplexity is really good imo, i usually get a niche answer much faster than skimming through yt vids
@MichaelStolarzthere is no week
@MichaelStolarztrue for mail; i just remember the checked-most-recent number of emails. it’s 230 for me. so if i see 232, i know i have 2 new emails to check
i need to fix this, typing this out on x made me realize how retarded this makes me
thanks twitter void, brb
@dannypostma
I don't get inbox zero. Just ignore stuff, it's ok. You're not required to reply.
@MichaelStolarz"Wait, a new meme format?"
"Well Joe, you've got different memes for different people. See, there's this thing about memes - and I know because of Baron. Anyway, he and this thing called 'teapot' on Twitter? They make memes, and Joe, you should see the money they make on 'em."
@MichaelStolarz(oh look, another shitting on apple post from me) the thing i hate most is that when apple products do the extraordinary, IT'S EXTRAORDINARY (duh). but when it glitches/fucks up (i'm looking at you, airpods packet drops), the magic behind it starts to have a bitter aftertaste
@MichaelStolarzi'm really phasing out of building apps in swift, it's just not worth it, i just want to do everything in console, not be forced to fuck around in xcode. idk today im really hating on apple even though all my devices are apple. maybe i'll switch for a clean 2025, idk
@MichaelStolarzwait a second, facebook has a dating feature? this is the first time i've ever heard about it. i can't fathom the quality
@MichaelStolarzhow can apple's music app - something that is part of their backbone as a company - be so completely and utterly shit? same goes for the new look the photos app has now. the only thing keeping me using apple music is the spatial audio. that's it.
@MichaelStolarzimagine if @RLGRIME doesn't drop a halloween mixtape this year? this would end me.
@MichaelStolarzlol monks using notion for meditation tracking. honestly kinda based tho, imagine achieving enlightenment and logging it in a kanban board
@IAmPascio
Ivan tells us that he recently was on a call with a special Notion customer: 2 monks from Thailand. The monks told him that it's the "everything toolbar" for their life and business, from meditations to events. When he started building Notion years ago, he would've never imagi… Less
Ivan tells us that he recently was on a call with a special Notion customer: 2 monks from Thailand. The monks told him that it's the "everything toolbar" for their life and business, from meditations to events. When he started building Notion years ago, he would've never imagined that.
@MichaelStolarzArc has a real chance to position itself as the emacs of the current era. People can live 100% in their Arc browser, all apps communicate and share data with each other to boost personalization/efficiency, all done through Arc and (hopefully) private on-device data analysis
@avstorm
The browser company is cooking something new
@MichaelStolarzsometimes i just don't get it. i fought with the nurses to let me in to see my baby's first ultrasound, i said idgaf how much i have to pay to have a "private room" with my wife during child birth so i could be there with her, never did it cross my mind to... GO HOME???
@MichaelStolarzsometimes i think about making my code more "production ready" but then i remember sonnet is the only one who's ever gonna read this garbage anyway
@MichaelStolarzit'd be cool to see something release today from openai as a response to the new sonnet; rn is their time window of release, so I guess we'll see what happens within the next 30 minutes from now. though tbh, even if they released a full o1, i'd still use sonnet (cost/speed)
@MichaelStolarzpeople posting "is the new sonnet actually smarter?" nah bro - they faked all the benchmarks, paid off paul to post fake aider new SotA results, everyone on tpot is paid off to lie about it positively - smh YES it's better, not opus 3.5 levels better, but better.
@MichaelStolarzreminder:
this is the worst this technology will ever be. and now it's in the hands of the likes of you and me.
@thedorbrothers
Makin' Trouble Song: Truffles by @mickjenkins @elonmusk @lexfridman @ianmiles @PicturesFoIder @cb_doge @AutismCapital @tunguz @kirawontmiss @charliekirk11 @FearedBuck @DefiantLs @TheFigen_ @WallStreetSilv @Rainmaker1973 @liz_churchill10
@MichaelStolarzthe old sonnet 3.5 was the autistic nerd that you bullied into doing your homework, apologized all the time, said "Certainly!" way too much; the new sonnet 3.5 is the chad genius who just solves your problem without any small talk or explanation
@MichaelStolarzbeing alive during the beginning of this openai vs. anthropic war of power brewing is so fun. this past year so far has been just the firestarter. imagine what they'll be competing with in 5 years time from now?
@MichaelStolarzHOLY FUCK I THOUGHT IT WAS JUST ME
i'm the one on the left btw
@MichaelStolarzcoming soon to obsidian (@obsdmd) x systemsculpt ai plugin :) the way that @kepano made canvas is so intuitive, easy to understand, and so comfortable to mod on. lots of work for me left to do before a full release but it's looking like sometime next week!
@MichaelStolarzgithub copilot seems to be just as useless as it was back in 2022. the only useful feature it has is generating commit messages (for some reason @cursor_ai never answers why they don't implement their own little "Generate Commit" button). anyway, back to Cursor/Aider.
@MichaelStolarzDOM-use, inspector-use, terminal-use, API-use all would be much better.
@dejavucoder
computer-use honestly doesn't seem to be a very smart direction. we probably need to build separate interfaces for llm agents.
@MichaelStolarzx's API is the most dogshit API I have ever come across to this date. v1? v2? oauth? client id? client secret? api key? oauth2? bearer? access token? api secret? token secret?
@MichaelStolarzif they released this new sonnet as opus 3.5, it’d be a little underwhelming BUT people would still pay the premium for the extra brain power it has over the original sonnet 3.5
@MichaelStolarzdamn, this new sonnet 3.5 got hands
@MichaelStolarzit's gonna be another one of those sleepless night coding with a new SotA model, DAMN it feels good.
@MichaelStolarzanthropic's new sonnet 3.5 implemented a new text editor specifically for @paulgauthier's Aider.
it's so over you have no idea, we need new benchmarks, this will probably hit 100%/100%.