builder, ai, markets, and general data nerd | https://xitter.cf/t.co/zs1Hb6u5e0

In the lab 🧪
Joined March 2025
Winston B. retweeted
Tough day for folks (DM me if I can help!) but if I found myself suddenly laid off from a company that cited AI as a cause, this is what I’d do: - download codex and Claude/code - say: “this was my job and how I spent my day, how can you help me automate it w skills” - push a dozen of those skills to GitHub - open up Claude design and make a portfolio site, with an “agent” per skill explaining how you built it, what tools it interacts with, etc.” give to lovable or v0 or whatever to publish - post that site and link to GitHub on LinkedIn - search “ai for ” and try all the new startups, form an opinion, message their founders - try something scary like openclaw, form an opinion - take a course in tactical AI in your field - build, share, build The gap in AI adoption is getting bigger. Start reskilling now while it’s early. The time is now.
55
130
30
2,002
154,837
Winston B. retweeted
Consumer sentiment is at historic lows. Unemployment is 4.3%. 63% of Americans rate the economy as bad while the labor market runs near full employment. This hasn't happened before at this scale. The jobs data is accurate. The sentiment is also accurate. People are describing different things: one is a statistic, the other is a cost-of-living verdict. axios.com/2026/04/13/america…
1
1
1
61
Winston B. retweeted
The ultimate "they don't know" alpha leak here 👇
Replying to @garrytan
Yes, and I think the skill gap is way larger than most people admit. For heavy users, we're finding that we can encode entire workflows to run night and day on plain english via markdown. Engineering context = engineering code in today's era.
2
4
103
36,301
Winston B. retweeted
Replying to @Hesamation
BRO. TrumpetMac and SlapMac being released within such a short time frame only means one thing is next 😭😭😭
1
187
Winston B. retweeted
🚨 Tin-foil hat rant: If you think about the timing of this release, you could argue this was deliberate. > OpenAI is playing catchup with their superapp. > Gemini with new models and features. > Meta with their new Brain algo. And then Anthropic stamping on top of it, further cementing the lead with a new frontier model "leak". Wonder how Capybara will perform at release though.
Anthropic’s new model, Capybara: “Compared to Claude Opus 4.6, Capybara achieves dramatically higher scores in software coding, academic reasoning, and cybersecurity.” According to Dario's previous interview, it might be a 10T-parameter model that cost $10 billion to train.
1
1
98
Winston B. retweeted
What it feels like to change thinking effort in Claude Code from High → Max
1
1
100
Winston B. retweeted
Let's talk about it connect!
My proposal for Meek Mill: I’ll fly you to Austin (private jet + 5-star everything). My team of instructors will train you in AI personally. We rally with the team putting AI in classrooms for thousands with Texas Sports Academy (free in Texas w vouchers). I’m 100% serious.
71
49
14
1,379
284,257
Winston B. retweeted
i set up Telegram as a remote control for my marketing agents in 15 minutes... and now i send tasks from my phone while i'm out the setup: > create a Telegram bot through BotFather (60 seconds) > install the @anthropic/claude-code-telegram plugin > add your bot token to the config > launch with: claude --channels > pair your account through the link it gives you you text the bot "write the full email sequence for our upcoming launch" and it runs on your local machine with full access to all your project files and context i've been running this from coffee shops, on walks, between calls... the machine at home just executes (and more reliably than OpenClaw)
35
16
1
220
16,860
Winston B. retweeted
Elon, Andreesen, Sacks, Ackman, Chamath, Vivek, on and on, I remember when for all these guys it was critical to elect Trump because our national debt was a crisis and existential threat to the future of America only he would tackle. Don’t hear so much about that anymore.
187
938
62
8,285
266,126
Winston B. retweeted
So you're telling me that on the brink of WW3, agents replacing people in the workforce, and biological cells learning to play Doom, a Chinese student hand-coded an open source project that: → Takes any scenario → Spawns millions of agents → Then uses those agents to simulate possibilities of outcomes and futures And then got the bag to the tune of $30M for this. AND WE'RE STILL TALKING ABOUT THE US GOV VS ANTHROPIC
20岁大四学生,用10天Vibe Coding手搓出来的开源项目:MiroFish,连续3天霸占GitHub Trending榜单前列,目前Star数已狂飙至8.3k,且还在持续增长中。 盛大创始人陈天桥看了Demo后,24小时内当场拍板注资3000万。 它是一个「预测万物」的 AI 仿真沙盘。只要丢给它一段新闻、政策或金融信号,它会自动生成成千上万个具备独立人格的 AI Agent,直接推演出未来的发展方向。 当你还在靠人工收集数据、靠经验做判断时,此项目已经开始用成千上万个 agent替自己预演未来了。 仓库地址:github.com/666ghj/MiroFish
1
3
148
Winston B. retweeted
Replying to @Polymarket
POV you just bought the dip and see Cramer's post
1
1
33
1,219
Winston B. retweeted
Replying to @ns123abc
Imagine training for real combat and to save real lives, just to get deployed to defend a data center
21
118
5
2,506
69,492
Winston B. retweeted
Replying to @om_patel5
First model about to be carnitas-guac-pro-3-preview
13
36
3
4,097
313,871
Winston B. retweeted
🔥🔥 MVP WORKING 🔥🔥 I built a device that gives me AI on any computer I touch. No software. No installs. No logging into a platform. No permissions. Plug in one USB cable and go. > Speaks into any text field like a keyboard. Works on Windows, Mac, iPad, anything. > Runs Whisper + LFM for speech-to-text and cleanup entirely on-device. No internet. No cloud. > Has Claude Code on-board so I can bring AI with me wherever. Recording a full demo soon. Excited to share more.
9
2
1
28
1,499
After another condescending model and a condescending reply from the CISO of OpenAI, I have permanently and irrevocably deleted my OpenAI account. For context, I was using GPT-2 back when they were still open source, training GPT-2 to unscramble sentences. Then I was on their forums teaching people to finetune GPT-3. Building cognitive architectures before "agents" were a thing. This company is rotten from the top, and they are not improving. I am done. Enough is enough.
He deleted this tweet but I kept receipts. Shitty behavior from the C-Suite at OpenAI. They are really hurting from the QuitGPT movement and I'm seriously about to delete my account.
69
101
23
1,005
131,006
Winston B. retweeted
Today ggml.ai joins Hugging Face Together we will continue to build ggml, make llama.cpp more accessible and empower the open-source community. Our joint mission is to make local AI easy and efficient to use by everyone on their own hardware.
I've started a company: ggml.ai From a fun side project just a few months ago, ggml has now become a useful library and framework for machine learning with a great open-source community
142
233
68
1,644
313,560
Winston B. retweeted
If you didn't have time to read the essay, I made an investment tracker of equities across the AI value chain. Explore nine layers of the AI value chain by market capitalization and build your own portfolio, performance benchmarked to SPY. Browse interactive charts across raw materials to processors to foundation models. intelligencetrade.vercel.app… Not fully comprehensive, not financial advice.
21
17
3
354
54,365
Winston B. retweeted
Yall just don’t understand how skills are also a game changer for chatbots as well now. We need an in depth article RIGHT NOW PLEASE @AnthropicAI @claudeai
1
3
128
Winston B. retweeted
Stop giving all your data to big AI companies. In our latest prerelease of Copilot for Obsidian, the agent runs without internet using @lmstudio . You can also self-host any model with native tool calling support and use it in Copilot with self-host mode. This is the first step of our "file over app but for AI" mission. We have a lot more to share soon!
7
4
1
64
4,082
Okay, I hear you. Really just trying to square this because based on my Cursor usage of the last 30 days vs. Opus 4.5 usage, Opus 4.5 is at least 2x more expensive for me based on avg. cost / turn. My codebase is fairly large with multiple services, but even when Opus 4.5 used fewer tokens per turn, it still costs substantially higher than my GPT calls.
1
1
78