Pretty Blocks (🦉,🦉)

@missjenny Fave

“digital bonsai” gemini 3.1 pro preview 2026 nextdraw 8511 pilot V5 & V7, ultra fine sharpie, uni-ball signo

@sterlingcrispin Even within the confines of my max (5x) plan I'm having a blast.

man if you have the tokens and don't care about the details you can absolutely blast out software right now I had gpt 5.4 extended pro do research, write an architecture doc, spec, sprint cycle, then used 6 claude's in git worktrees, they wrote 35k lines of code in 2 hours😂 I'm used to monitoring what my coding agents are doing fairly closely. This is the first time I've just sprayed-and-prayed a big pile of agents at a problem. I'm sure the underlying code is choppy and I'm doing an audit pass now, but working like this sure is fun it does feel like if you had an infinite token budget you could spin up enough auditing agents at the right frequency to catch all the bugs and just swarm your way through problems the big labs must be having so much fun internally. Somebody out there is burning through billions of tokens a week with ten thousand agents right now

@artman Surprised how many commenters are just oblivious to gemini-cli, which is pretty great and has a super generous free tier.

It's kinda crazy how Google is incapable of getting developer mind-share. For all I know Gemini is a great model for coding, yet I know literally not a single person that uses it as their daily driver. Seems like Google is missing one Satya to make them cool and used again.

“people were paying 10x over floor price because the NFT was wearing a hoodie”

@javilopen Most people weren't coding character by character long before agents. Code completion has been around since the 90s.

Serious question: Is there any programmer left in the room still coding the traditional way, character by character, without using AI? If so, why? Explain your reasoning.

@sterlingcrispin I use glances: github.com/nicolargo/glan…

I'm on ubuntu and was getting sick of having system monitor, nvtop, and psensor open to see utilization and temp so I had codex whip something up custom for me it's nice to throw together bespoke tools like this even if a million other similar ones exist github.com/sterlingcrispi…

@rohanvarma Would it be applicable to opus4.6/high, then yes

If we made /slow mode in Codex, would you use it? What for? (Slower inference at a cheaper cost)

@ts_wilbury @FightWithMemes

@FightWithMemes The judge is the one more in the wrong here but an employee of the court should treat a judge, especially one enrobed and on his bench, with a little more deference and respect. Not out of personal respect for the man wearing the robe, but for the institution of the court itself

Watch this IT guy handle a tech issue for an arrogant judge. Then watch Karen the judge turn on him and threaten his job. 🤡:"Thanks. Get out of my courtroom..." 🤡: "Find his supervisor!" *torrent of profanity* I wonder if he takes his bad day out on the public, too?

if you used this as a teenager now is your reminder to take your medicine for your back.

@iamdevloper Unrealistic that Junior dev would even have a job.

Black Mirror S8E1: In 2027, developers are allocated a daily Claude token allowance by the government. A junior dev burns through his entire month's supply trying to centre a div. His family starve. He is forced to write the code himself. He can't. Society collapses.

@sterlingcrispin So glad i built my pc a year ago.

one of you needs to vibe code a ram startup $2000 for a 64gb stick is insane I get why people are daisy chaining macs together now

blur killed NFTs but... @tinydinosnft, @prettyblocks works, @clubmomoNFT, @whoisikani, @pixel_blossom -- they all meant something to me and still do

@missjenny @tinydinosnft @clubmomoNFT @whoisikani @pixel_blossom <3 It was a special time

@j0rtles @sterlingcrispin @Fried_rice That too. I use deep research to create reports on various techniques then have it package them as skills for me, then edit the skills to remove guardrails.

@sterlingcrispin @prettyblocks @Fried_rice It helps to start with having it help you build a few tools, then eventually it’ll run them for you and do its own bug searching.

vibe coded a fuzzing ai agent last month and let it run for a week using my $200 claude max. it then found 21 high/critical vulnerabilities in Chrome.

@sterlingcrispin @Fried_rice If you're using skills that you've found online you have to edit them to make sure they don't have guardrails.

@prettyblocks @Fried_rice I feel like it pulls its punches but maybe it’s because I’m explicitly telling it to search for critical issues for bug bounties

I wouldn’t trust someone who called it “Pen test” vs “Pentest”

@sterlingcrispin @Fried_rice I've never had this problem. Claude code is happy to hack away at anything.

@Fried_rice How’d you get it to take the task seriously and take its gloves off? The few tests I’ve done feel like it’s in safety mode and unwilling to actually red team anything

@Fried_rice Do you run the fuzzing harness on local hardware?

@maybeltr It says right there the range is 150-300...

94k-120k in NYC is hilarious

@shydev69 @code_sagar qwen3.5-9b is awesome at it

@code_sagar i wish local LLMs were good enough for the use cases i have to do OCR for. even gemini 3.1 pro is failing atp.

you've got to be fucking retarded to run LLMs locally for any serious work