gabe_404

@luisocampox @AnthropicAI lets goooo!!! excited to have you on the team!

life update: after nearly 5 years at partiful, i’ve joined @anthropicai to manage our influencer community and events pumped to be a part of this incredible team. if you’re a creator who’s been building with claude, i’d love to chat :)

bring back shaming actors, musicians, and celebrities for selling out for advertising money

@michaelmiraflor Spirit of the Machine, hear my prayer, Guide this missile, hold it true, Let it part their steel and weak armour, And crack their cowardly skin, And smite the foe from the Omnissiah's sight.

New age of spiritual warfare is upon us. Archangels among us. Holy swords to slash through the slop.

@redbull aderall

what’s the best combo with a Red Bull?

is it just me or does the YC brand work negatively for most founders and builders in top techcircles? I would never trust, hire or work with anyone from YC class its just kind of a big grift at this point or is there something im not seeing?

@hopes_revenge its giving

> mom found the data center water jar

@chamath shilling 8090 like his entire portfolio depends on it

Microsoft pulling Claude is the first, but not the last. The issue isn't that the tool isn't useful. The issue is that without context and oversight, the tool can spin forever and generates an enormous cost burden that, when cascaded across an entire employee population, makes using the tool economically untenable. 8090's Software Factory is the control plane that is becoming increasingly used by Enterprises to get the job done but do it in a smart and scaleable way.

@tryramp was one of the top brands in @milkkarten Link in Bio survey on what account exemplifies great social. so i ran our audience intelligence tool on their Twitter account to understand their audience we started from Ramp's X account and mapped over 19 million relationships from who their followers followed. we identified over relevant 8,400 accounts their audience follows, 47 communities and 6 unique audience personas here's what their audience actually like: Frontier ML Builders — deep learning researchers, open-source model contributors, and agent infrastructure developers. anchored by Clem Delangue and François Chollet. analytical, fiercely curious, and treating rationalist philosophy and weekend coding experiments as part of the same pursuit Hard Money Populists — alternative market strategies, hard money convictions, and populist political commentary blurred into a unified worldview. Robert F. Kennedy Jr and zerohedge set the daily tone. off-screen it shows up in ancestral health, solitary hikes, and self-sufficiency Indie Interface Crafters — component-driven design, interactive web interfaces, and bootstrapped software businesses. following Emil Kowalski for animation inspiration and shadcn for reusable components. building profitable solo products instead of chasing venture funding Blue-Chip Macro Realists — major political developments, enterprise technology, and mainstream sports sitting side by side as daily public record. The New York Times and LeBron James anchor a worldview that tracks globally established institutions and icons Rationalist Venture Backers — early-stage investing, industrial manufacturing, and rationalist philosophy. Bill Gurley on venture mechanics, Trae Stephens channeling capital into defense and physical infrastructure. personal optimization and dense economic texts fill the downtime Onchain Protocol Technicians — token speculation, smart contract security, and blockchain engineering. Toly building high-speed infrastructure while samczsun hunts vulnerabilities in the code. a tense mix of rigorous engineering and sharp financial nerve what stood out: Ramp is a finance platform but their audience isn't "finance people." it's six distinct cultural communities that each relate to money, technology, and independence in fundamentally different ways. a single content strategy aimed at "finance professionals" would miss most of them the full report with all 47 communities, persona breakdowns, messaging recommendations, and the accounts inside each cluster is explorable here: lnkd.in/gAiDpc8u reach out with any questions! ~ gabe {ⓛ ω ⓛ}

@nikitabier Marlboro Reds, Diet coke, and an nvidia H200

Next week I’m headed to Tokyo to meet with the X Japan team. What is the most American thing I can bring them?

today's date is dynamically placed into the system prompt, and the models have to do calendar math from there to figure out any other day. they're not great at that math, and it degrades over longer convos. just telling it the date usually fixes it. also helps to use the newer reasoning models (GPT-5.5, Opus 4.7) with extended thinking, they will actually work the math out instead of guessing (the system prompt is a hidden set of instructions the model sees before your first message)

@cjc the psychographic trait you are referring too is "need for cognition"

I firmly believe developers don’t want selfie sticks and a “Get Ready With Me” dev edition. Everyone is fighting for attention now, even the giants. But developers are a unique audience: they are discerning with a high bullshit detector. When you replace technical depth with selfie sticks, you very well might get the reach, but you lose the earnestness that developers actually value.

here is the pipeline I would build 1) get top 20 YouTube Videos 2) collect all there comments 3) gemini API to transcribe, describe, and tag each video, and the topics covered 4) embed all the comments 5) dedupe and near dedupe comments 6) filter for low quality comments 7) tag all comments with Gemini Flash, using the video description and transcription as context 8) run sentiment analysis on all comments, using the video description and transcription as context 9) run HBDSCAN clustering on the comments 10) check that they are actually stable and valid 11) for each cluster, give the comments of the cluster to the GPT 5.5 to have it name and describe the cluster 12) examine the results :)

@rileybrown bro really got scared after 1000 comments?

bruh

People, especially young people hate AI.... Let's try and figure out why.

“Ok, Google IO is about to start. Have Karpathy announce he’s joining Anthropic.”

@karpathy bro really needed those unlimited tokens

Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.

@yagiznizipli please

What issues/bugs do you see on x.com? Share with us and I’ll get it fixed.

@jaiyagill just caffeinate your mac

people at the frontier of technology are carrying around half-open macbooks or inventing hardware dongles because they somehow don’t know “keep mac awake” apps already exist

@ctatedev can someone please explain how this is better then just using rust or python, which models have enormous training on?

Introducing Zero The programming language for agents. I wanted a systems language that was faster, smaller, and easier for agents to use and repair. Explicit capabilities. JSON diagnostics. Typed safe fixes. Made for agents on day zero.

the whole idea of making an "agent's first coding language" fundamentally flawed because models are good at coding because of the massive training corpus either the new language is genuinely novel (in which case agents will be bad at it because no training data) or it's just a remix of existing languages (in which case why not use the originals the models are actually fluent in instead of making them guess which leads to degraded outputs?)

I have reviewed the language and really tried to understand this, but I really do not understand this language's purpose other than engineers with too much free time, free tokens, and a marketing budget. I was very excited to read about a language is "agent's first." Its just zig with a touch of java and rust...

leaving sf has become the equivalent of touching grass