I probably make cool games (over 235M+ visits!) Very into ML/AI
Joined September 2025
- Tweets2.3K
- Following671
- Followers467
- Likes1.9K
this is what I expect a model pre-trained on everyone who lives in SF to sound like
i've lived in sf for well over 10 weeks
seen everything
startup founders come and go
kids get accepted into YC and then get chewed up by this town in 3 days
i've dropped out of YC, thiel fellowship, and a stanford admissions tour
now i'm building an AI agentic AI startup in the AI space
already raised $50 million seed round in a dream I had last night
i'm the next sam altman and dario combined
hi @chooi_jeq I was wondering if you guys benchmarked deepseek v4.1 1017 and/or any other open source model on your benchmarks? Would love to see some results on them!
Where's my cash twin (also in the process of trying to figure me out, it doxxed a random guy with the same name and found his password, wtf lmao)
We partnered with @AgentPhoneHQ to give our AI agents $10,000 to hand out to everyone watching this.
Comment “agent” under this post, and we’ll text you a number.
Sixtyfour finds the fun facts about you from your digital footprint, and Agentphone talks to you.
If we get any facts wrong, you get paid.
Comment “agent” now. You have 48 hours!
SquaredCube retweeted
Agreed - it's going to be similar to TikTok / YouTube where everyone 'can' do it, but actually being 'good' at it is something else.
We are 100% going to see some amazing breakouts though from people who have never done any dev work ever - exciting (and daunting) times ahead
oh my god.
i wake up to this????
le chaton fat was NOT fake?
Meet Mistral Large 4, aka Le Chonk.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
Roblox just rolled out a wave of the Crowns of O's after previously discontinuing it!
Who got theirs?
Yoo I finally got mine
W @Roblox
SquaredCube retweeted
To celebrate 50k followers...
I'm giving away 3 Viridian Domino Crowns! 💚
To Enter:
Follow @LukasThePickle
Like & Repost
Picking the winners on Sunday! 👑
this is a really good reaction gif btw
Introducing Diffusion Controller, a lightweight steering damper that precisely steers image generation for better prompt alignment without breaking stability. Read the blog to learn how it boosts image quality without breaking baseline stability → goo.gle/4hFCZiW
hi @twilio
I won this & filled out the form you guys sent, but got 0 update since, its been a few weeks (and I live like 1 hr from your office in SF)
any updates on this?
Every builder has a bug story. Today, we want yours. 👀
In honor of Programmer's Day, drop a comment with the weirdest bug you've ever spent way too long fixing for a chance to win a Twilio-branded mechanical keyboard.
Submissions to enter drawing will be accepted until midnight EST on September 14.
Winners will be announced via DM on September 15 EST.
SquaredCube retweeted
iBUYPOWER Masters #FML2026 Finals are here and we have a special announcement...
We're giving away ONE(1) @AMDRyzen 7 9800X3D and FML2 Hoodie to one lucky winner!
Enter for your chance to win: ibp.gg/FML-Giveaway
Like + RT for extra entries! | NA Only
SquaredCube retweeted
Seeing ICLR 2027 has 60000 submissions:
The conference is designed to expose research ideas and findings or even breakthroughs. Under today's conference volume and super noisy review system, that purpose is 99% gone. Most papers will just be buried among thousands of papers, and few people can actually read them.
In this context, the right mindset is to optimize for impact instead of publication. The impact can be a super solid and inspiring paper, a metodology that can be adopted at large scale, an excellent demo and project page, a well-maintained codebase, or a viral Twitter post that many big names are sharing. Don't get emotionally attached to the conference decisions, SIFT/DiT/LSTM/Disillation were rejected even multiple times, and Transformer/GAN/Adam or more recent Dust3R/JiT turned out to be just a poster paper but yet had huge impact.
Just keep doing the right thing and optimizing the impact 🙂
