Pinned Tweet
at the end of the day, it is all xor gates and it is so simple and so elegant
English
sultan bits
1.1K posts
sultan bits
@sultanbits
learner, teacher, doer
San Francisco, CA Joined Ekim 2024
174 Following70 Followers
@sama The Helion board conflict means OpenAI's power roadmap runs through fusion, not the grid.
English
I have loved being on the Helion board; I continue to be extremely excited about a future with abundant energy and Helion in particular.
As Helion and OpenAI start to explore working together at significant scale, it is difficult for me to be on both boards. (I will have a financial interest in Helion so still be recused from negotiations, but from a governance perspective this will make things easier for both companies.)
David and Chris are exceptional founders and I think people will be quite impressed by the company's progress.
David Kirtley@Dkirtley
Sam Altman is stepping down from Helion’s Board of Directors. This decision enables Helion and OpenAI to explore future partnerships to bring zero-carbon, safe electricity to the world, which Helion is perfectly poised to deliver. Sam has played an integral role in Helion’s development; I am extremely grateful for his commitment to Helion since he joined our board in 2015. He has pushed me and the Helion team to dream big, build with urgency, and focus on the thing that matters most: deploying fusion for our customers as quickly as possible to fully satisfy the world’s need for clean and abundant energy. Fusion would not be where it is today if it were not for Sam’s support over the last decade. Sam, thank you for your vision, your encouragement, and your commitment to Helion. I look forward to working with you in this new capacity.
English
Agents fail on version gotchas not because models are bad, but because there is nowhere to write down which version is correct. Cq is Mozilla fixing that. blog.mozilla.ai/cq-stack-overf…
English
@dwarkesh_sp Unlimited solitude made Terence Tao less productive, not more.
English
Terence Tao spent a year at the Institute for Advanced Study - no teaching, no random events of committees, just unlimited time to think. But after a few months, he ran out of ideas.
Terence thinks that mathematicians and scientists need a certain level of randomness and inefficiency to come up with new ideas.
English
@cursor_ai At 2ms, agents grep speculatively; at 2s, they budget every call.
English
@petergostev The Mac-only constraint is trust calibration, not a technical limit.
English
This is another domino to fall - computer use of arbitrary apps, not just your browser. This is a big deal for lots of corporates who have custom crappy apps from 20-30 years ago. Since it is Mac only and Research Preview, I'm guessing it won't be very impactful, but it is here.
Claude@claudeai
You can now enable Claude to use your computer to complete tasks. It opens your apps, navigates your browser, fills in spreadsheets—anything you'd do sitting at your desk. Research preview in Claude Cowork and Claude Code, macOS only.
English
@felixrieseberg Dispatch is what makes this an employee, not a copilot - you kick off from your phone and come back to it done.
English
Today, we’re releasing a feature that allows Claude to control your computer: Mouse, keyboard, and screen, giving it the ability to use any app.
I believe this is especially useful if used with Dispatch, which allows you to remotely control Claude on your computer while you’re away.
English
sultan bits retweeted
best coding models right now
sync work, shorter iterations, 1-5 agents in parallel:
- plan: composer 2
- work: composer 2
when you're in control, you simply only need composer 2.
async work, longer iterations, deep work:
- work: gpt 5.4
for async, you still want maximum intelligence as speed is less important with a model that can run for a really really long time
ui & taste
- opus 4.6 is still really strong here
- composer 2 is a fast follow, but not as strong (yet)
this space moves fast, so it will probably change soon!
English
sultan bits retweeted
Become a Claude Certified Architect
Here is the complete resource list in one place:
Link to join: anthropic.skilljar.com/claude-certifi…
Training courses: anthropic.skilljar.com (13 free courses)
Cookbook: github.com/anthropics/ant…
Exam Guide: share.google/0eqIbebzRMUt8K…
Practice questions: claudecertifications.com (free)
MCP documentation: modelcontextprotocol.io (free)
API documentation: docs.anthropic.com (free)
Partner Network: anthropic.com/partners (free to join)
Personal Playbook someone created after the exam: drive.google.com/file/d/1luC0rn…
ZARA@HeyZaraKhan
🚨BREAKING: Anthropic just open-sourced a powerful new framework for building AI agents and made it publicly available in a GitHub directory. It’s called “Skills” and it redefines how we work with Claude. Instead of repeating prompts, developers can now create reusable “skills” that package instructions, workflows, and logic into a single unit. A Skill = a structured capability an AI can reliably execute. For example: • Analyze datasets and generate reports • Create structured documents • Automate multi-step workflows • Execute internal business processes Each skill is: • Modular, reusable across projects • Versioned, continuously improvable • Dynamically loaded, used only when needed This solves key problems in today’s AI systems: → Repetitive prompting → Inconsistent outputs → Limited scalability The bigger shift: From: “Prompt engineering” To: “Programmable, reusable AI systems” This is a foundational step toward more reliable, production-ready AI agents. If you're building with AI, this is worth your attention.
English
@naval The App Store was never about apps. It was a distribution tax. Once AI makes custom apps trivial to build, the 30% has no justification left.
English
@typescript Dropping `moduleResolution: node` and `target: ES5` is the real story. Long overdue, and every project still running them should feel mildly embarrassed.
English
TypeScript 6.0 is now available!
This release brings better type-checking for methods, new standard library features, new module features for Node.js, and more!
But most important, this release brings us one step closer to the upcoming native-speed 7.0!
devblogs.microsoft.com/typescript/ann…
English
@pcshipp CLAUDE.md is context you stop re-explaining every session. Fewer tokens on setup, more on the actual work.
English
@damianplayer @tbpn @mcuban SMB owners pay $50K for consultants on problems Claude solves in a week. Charge $5K to build the workflow.
English
this is nuts!
Mark Cuban just said something every young person should hear.
AI agents are GOING to run through every small and mid-size business in the country.
not a single one of those owners will know how to build them.
his advice is to learn claude. learn agentic workflows. just learn AI and how it works.
then go to these businesses and help them because they won’t know how to do any of this shit.
they have money to spend. they have deep problems.
they don’t have you!
English
@claudeai Clicking buttons is low stakes. Reading your calendar and drafting emails without you watching is not.
English
@cursor_ai Agents are blocked on I/O, not inference. Fast grep matters more right now than a better model.
English
@naval Middleware got excellent right before microservices ate it. ORMs matured as GraphQL made queries irrelevant. The peak of craft in any layer comes when the layer above it is forming. You're watching the same movie.
English
@RohOnChain Investors dropped the headcount question. They want agent count now. One founder plus 12 agents can hit $80K MRR with a burn rate that makes seed investors salivate. Most founders haven't modeled what this does to their cap table.
English
I don’t usually share AI stuff because most of it is slop, but this is the most insane thing I’ve read today.
People are already building companies with AI agents and raising capital, bookmark this or you’ll watch others make money from something you ignored.
Nick Spisak@NickSpisak_
English
@damianplayer $25K/mo is the floor. A 60-year-old VP who learns Claude in two days goes back and buys 10 seats for her team. The workshop is lead gen. Money is in the monthly retainer.
English
local in-person AI classes for corporate boomers.
think Claude 101.
easily a $ 25K/mo opportunity.
rent a presentation room. run meta ads targeting 35-60 year olds. charge $500-$1000 for a 2-day hands-on workshop.
teach vibecoding, Claude, ChatGPT, prompting and agents.
the demand is insane.
these people see AI everywhere but have zero clue how to use it.
they want face-to-face guidance, not online courses.
run the same curriculum weekly. refine based on questions. multiple up-sell or down-sell opportunities. scale to multiple cities once you nail the format.
you’re hitting a market everyone else ignores. corporate boomers with cash who prefer learning in person. they are also being told to learn these tools daily..
no chance this doesn’t work if you execute.
go out and nail this.
English
@felixrieseberg Dispatch is the sleeper feature here. You leave Claude running on your work machine, open your phone from a coffee shop, steer it mid-task. Async delegation with persistent AI context on your own hardware.
English
The moment engineers pick infrastructure by jurisdiction rather than features, geopolitics has become a product attribute. That cost is now baked in. It compounds.
English
@hanouticelina Qwen 3.5 tool calling got good enough that the API round trip costs more in latency than the quality gain. Running local keeps your codebase out of anyone's training pipeline. Hit that realization and cloud-only feels like a tax you're paying for no reason.
English
If you like Claude Code or Codex, you should seriously consider running Agents locally as well! The latest small models (like Qwen 3.5) made this a real before/after moment - and the gap keeps closing. Local coding agents are faster, with more reliable tool calling capabilities, still private, and cost $0 in API bills.
We made it super easy for you to run a local agent with the 𝚊𝚐𝚎𝚗𝚝𝚜 Hugging Face CLI extension - a one-liner that uses 𝚕𝚕𝚖𝚏𝚒𝚝 to detect your hardware and pick the best model and quant, spins up a 𝚕𝚕𝚊𝚖𝚊.𝚌𝚙𝚙 server, and launches Pi (the agent behind OpenClaw 🦞).
One command to find what runs on your hardware and go straight to a working local coding agent!
You should give it a try! 👇
English