YourAverageDev

@realsigridjin add plz

Reply here if you want to be part of Canadian builder group chat in X 🇨🇦

not an april fools, just a groundbreaking new ai model this new ai model can design + build AI chips human's couldn't create themselves using electromagnetic behaviour they're literally called 'alien structures' you describe the electromagnetic behaviour you want and it generates the physical structure that creates it... - model is called Heaviside and works 800,000x faster (not a typo) than trad methods. - the designs it generates uses shapes human brains hadn't ever thought of. - claude opus 4.6, chatgpt 5.4 both failed on the same task (actually made it worse lol) - use-cases: novel ai chip design (interconnects), scaling 5G/6G, beaming solar power to earth from space ( elon would fucking love this) - all of its training data was SIMULATED. for me this is the sickest part, they're targetting 1000 YEARs of simulated data by end of year finally a model that understands the world vs. llm words thats NOT a world model

@MilksandMatcha I need it mainly for ML training and try to get autonomous research on it working

Giving away 5 Codex Pro plans Each person will get 3 months of free Codex Pro (highest tier). Winners will be selected from comments in 48 hours, comment below why you want it.

@0xSero random

Giving away 5 Opencode Go subs Winners selected randomly from comments in 24 hours.

🚨 BREAKING: FEDERAL JUDGE JUST DESTROYED THE PENTAGON CASE AGAINST ANTHROPIC The Judge found Pentagon's own records show they designated Anthropic a supply chain risk for publicly criticizing the Pentagon's position >judge: "classic illegal First Amendment retaliation" >pentagon lawyers argued anthropic showed "subversive tendencies" by "questioning" and "raising concerns" about use of claude >the judge was not impressed >called the pentagon's argument "Orwellian" >pentagon claimed anthropic could remotely sabotage claude >judge asked: how? >anthropic submitted evidence they have ZERO access or control once deployed in classified systems >pentagon lawyer admits he was "unaware" of this >judge: so what's the legal basis for blacklisting anthropic from all contractors? >lawyer: "I don't know" >asked if Hegseth had authority for this: no >asked if it reflected DoW's actual intent: no >asked why he said it then: "I don't know" >emil michael wrote a memo calling anthropic a "hostile" "unacceptable national security threat" >NEXT DAY emailed dario: "After reviewing with our attorneys and seeing your last draft (thanks for being fast), I think we are very close here" >the judge: "It is exceedingly difficult to square this correspondence" >pentagon asked the court to impose a bond >the judge set it at $100 >that's the judicial equivalent of "get out of my courtroom" >the pentagon has 7 days to appeal The Judge found likely violations on the First Amendment, Fifth Amendment, and APA. Every single claim. ANTHROPIC OFFICIALLY WON.

@KeyTryer genuine question: how do you ban matrix multiplication, should we just stop teaching linear algebra in schools?

This is one of the most mentally ill subreddits lol

cursor composer 2 is officially out

we just can’t use backprop, the reason the models forget catastrophically is due to mainly backprop as its goal is to find the lowest loss on every new set of data, which means basically wiping out the weights from before and overwriting them. Transformers isn’t the problem modern deep learning is.

continual learning, only continual learning, and nothing other than continual learning, is what's missing right now I couldn't care less about saturating benchmarks, getting +3% in SWE Bench or whatever will not make these tools much better than they are, for as long as they still forget all they've learned in the next session AGENTS and MEMORY markdowns don't do it either, the amount of information in even a few hours of work is already several thousand tokens long, and any attempt at compressing that will either lead to important information loss, or overwhelm the context to the point that the model becomes brain dead please may the next launches be more about how this is fundamentally addressed and less about whether then can RL an LLM to beat ARC AGI 7 - which, by all means, is cool and impressive, but what everyone actually needs is an AI that doesn't feel like we have to onboard a fresh new intern every 2 hours of work

@krishnanrohit this has to parody / bait right?

Guess who's vibe coding in 2026 ... Linus Torvalds, using Google's Antigravity. It's over.

@MihalisBenko @Truthful_ast

@Truthful_ast SpaceX starship will absolutely never in a billion years ever go anywhere near the Moon. For a Starship to make it to the Moon it would need literal two dozen + other Starship launches to refuel the ship aimed at the moon. Type of refueling that was never even attempted.

Toyota’s Lunar Cruiser is absolutely massive, this will be the first ever real pressurized rover ever made and is capable of acting like a mobile base! Either SpaceX’s Starship or Blue Origin’s Blue Moon will deploy Lunar Cruiser on the surface of the moon. It’s the first time serious development has went into such a rover, NASA’s pressurized rovers never went as far as this

@allisx86 because shitcrosoft made it

I thought normies were finally into learning RL but apparently outside of tpot this means Ralph Lauren

@ericzakariasson wait this isn't a photo???? what the level of detail on cursor's website is insane!!!

@alxfazio Ghostty doesn’t?

been desperately trying to find a terminal that can actually handle claude code. kitty seems to be holding up best so far

@thdxr @TomZarebczan story? github.com/opencode-ai/op…

@FoundTheCode @TomZarebczan ?

there's plenty of stories of founders denying insane acquisition offers and it looks crazy from the outside but there's things you can only see from the inside that make it a lot easier to say no

@thdxr @TomZarebczan Didn’t you guys get acquired initially and it half fell through

@TomZarebczan lost count

@scaling01 Worse than Opus 4.5 but definitely way better than Opus 4 Was a game changer for me when 5 codex came out

I will die on the hill that GPT-5.1-Codex-Max and all other xhigh models are slop

"i actually think this is the worst algo iteration of all time and will lead a lot of power users to quit using the app"

@QuixiAI @XiaomiMiMo is this a joke or no

Make sure to disable Thinking when you use @XiaomiMiMo V2-Flash for vibecoding.