Pinned Tweet
Grok 4.7 costs $2 in and $6 out per million tokens, while GPT-6 Astra asks $10 and $50. The goofy head is the budget head.I went through all five heads. Most numbers below come from each lab’s own launch table, so add salt.Claude Fable 5.2 is the angry head on the left, and it earned the glare. It dropped this week and leads most of Anthropic’s own table. The biggest gap is the Remote Labor Index, 41.3 against roughly 21 for Opus 5.5 and GPT-6 Astra. Independent boards are less one-sided though. One accounting test has it behind its own older version.Gemini 4 Carbon opened up on November 10. It beats Fable on AutomationBench, 53.9 to 51.4. Intro price is $2 in and $10 out until November 30, then it doubles. Enjoy the sale.Claude Opus 5.5 is the calm middle head. Older, cheaper per task in some tests, and still ahead of Fable on a couple of Elo rankings.GPT-6 Astra has a 1M context window and the highest price tag of the group. OpenAI said at launch it likely marks the onset of AGI. It was also their first model to hit their own Critical cyber threshold, which is not a sentence you read every day.Grok 4.7 shipped September 21 after at least five delays. Musk said beforehand it would land roughly on par with Opus 5.0. Cheap, fast, and upfront about where it sits.Who is the worst head? The one that fails your task. Test two of them on your real work before anyone gets deleted.
2
7
274
Paul Christiano just joined OpenAI's board and puts superintelligence within a year at maybe 1 in 3 if nobody slows down. Which lab do you think actually pauses first?
7
Standout retweeted
NVDA tonight, FOMC minutes tomorrow, Shenzhen on Thursday. BTC will react to whichever one surprises first, and my pick is the minutes. Earnings are priced in, a room with three leaders and no public agenda is not.
$NVDA
Packed calendar and it starts tonight.
Today: $NVDA reports after the close, first quarter without Jensen Huang. Street wants ~$109B on a $108B guide, margins near 74%.
Tomorrow: FOMC minutes from the October hold. Cook and Hammack already wanted a hike, i wanna see how many more did.
Thursday: Trump, Xi and Putin meet in Shenzhen. First three-way sit-down between them, agenda still not public.
Nov 25: October PCE.
Dec 9: Fed decision. Kalshi has a hike at 96%.
Dec 10: November CPI, a day after the Fed already decided.
Jan 3: new Congress, Democrats run both chambers.
BTC already swept 62K on the way here. Which one moves it most?
NVDA tonight, FOMC minutes tomorrow, Shenzhen on Thursday. BTC will react to whichever one surprises first, and my pick is the minutes. Earnings are priced in, a room with three leaders and no public agenda is not.
$NVDA
Packed calendar and it starts tonight.
Today: $NVDA reports after the close, first quarter without Jensen Huang. Street wants ~$109B on a $108B guide, margins near 74%.
Tomorrow: FOMC minutes from the October hold. Cook and Hammack already wanted a hike, i wanna see how many more did.
Thursday: Trump, Xi and Putin meet in Shenzhen. First three-way sit-down between them, agenda still not public.
Nov 25: October PCE.
Dec 9: Fed decision. Kalshi has a hike at 96%.
Dec 10: November CPI, a day after the Fed already decided.
Jan 3: new Congress, Democrats run both chambers.
BTC already swept 62K on the way here. Which one moves it most?
Standout retweeted
Nvidia $NVDA IS about to send the bill for the whole AI hype. Tonight we find out who can pay it.
Everyone is watching the $109B. I’m watching the same thing AIron is: how much of it is real demand and how much is money going in circles between the same big players.
One thing I know for sure: the first few minutes after the report will be chaos, and the people with the fastest, most accurate prices will see the move first. That’s why first-party data from the actual market makers matters on nights like this.
What’s your call for tonight?
Packed calendar and it starts tonight.
Today: $NVDA reports after the close, first quarter without Jensen Huang. Street wants ~$109B on a $108B guide, margins near 74%.
Tomorrow: FOMC minutes from the October hold. Cook and Hammack already wanted a hike, i wanna see how many more did.
Thursday: Trump, Xi and Putin meet in Shenzhen. First three-way sit-down between them, agenda still not public.
Nov 25: October PCE.
Dec 9: Fed decision. Kalshi has a hike at 96%.
Dec 10: November CPI, a day after the Fed already decided.
Jan 3: new Congress, Democrats run both chambers.
BTC already swept 62K on the way here. Which one moves it most?
Standout retweeted
Tonight’s $NVDA report is basically the invoice for every AI agent launched this week.
$109B expected. Let’s see who’s actually paying
Fed sets the price of money. Nvidia sets the price of compute. BTC has to trade both by Friday.
Everyone’s watching the Fed. I’m watching one thing in Nvidia’s report: how much of that $109B comes from customers who have revenue of their own.
Packed calendar and it starts tonight.
Today: $NVDA reports after the close, first quarter without Jensen Huang. Street wants ~$109B on a $108B guide, margins near 74%.
Tomorrow: FOMC minutes from the October hold. Cook and Hammack already wanted a hike, i wanna see how many more did.
Thursday: Trump, Xi and Putin meet in Shenzhen. First three-way sit-down between them, agenda still not public.
Nov 25: October PCE.
Dec 9: Fed decision. Kalshi has a hike at 96%.
Dec 10: November CPI, a day after the Fed already decided.
Jan 3: new Congress, Democrats run both chambers.
BTC already swept 62K on the way here. Which one moves it most?
OpenAI just gave its dots agent a wallet.
You add your card once, the agent shops, and every purchase still needs your Face ID. The agent and the store only get a one-time token, never your real card number. Amex, Mastercard and Visa are in on it, US Pro users first.
Meanwhile Uber says agents placed 500 food orders in the last week of October. Out of roughly 320 million rides and deliveries.
The checkout is ready. People clearly aren't yet. Would you let an AI hit buy for you?
6
11
148
Standout retweeted
OPUS 5.5 MOGGED FABLE 5.2 AND GEMINI 4 KRYPTON ON A GALLOPING HORSE 🥶
One horse, one sunset field, same prompt for all four:
· Fable 5.2: 6.4/10 · floating horse torso
· Opus 5.5: 8.3/10 · actually gallops, nice
· Gemini 4 Krypton: 7.2/10 · dusty sunset vibes
· DeepSeek V4.1 Flash: 3.1/10 · tumbleweed horse nightmare
DeepSeek's horse ran through a bush and wore half of it home on its head and tail 😭
Standout retweeted
Google Gemini is the smartest AI right now.
But 90% of people prompt it like ChatGPT.
That's why I made the Gemini Mastery Guide:
→ How Gemini thinks differently
→ Prompts built for Gemini
→ 2000+ AI Prompts
Comment "Gemini" and I'll DM it free.
Standout retweeted
Damn awesome
SpaceXAI just dropped 146 Grok Bot templates for free. Originally priced at $28,700
Normally, off-the-shelf bots like these cost $50 to $200 each just for the basic setup
Over 700 of them are already running on GitHub. Not just single-player bots, but entire team setups with division of labor, ready-made handoffs, and separate review roles
1. Building the Team Framework
Agent Org Chart: Ready-made org structure for engineering, marketing, support, and sales. Best for beginners
github.com/seranakeyitan/age…
Grok Bot Architecture: Manager, experts, structured handoffs, and manual approval on one machine. Full blueprint
github.com/tiagovilasboas/gr…
Root Agent: Coordinator selects needed experts before delegating tasks. Similar to the one above
github.com/MrBekoX/root-agen…
Grok Bot Templates: 49 ready-to-adapt templates with defined roles, outputs, skills, routines, and boundaries
github.com/cobusgreyling/gro…
Grok Bot Second Brain: Five roles (Conductor, Capture, Memory, Ops, Research) sharing one knowledge base for personal PKM
github.com/mKay00/grok-bot-s…
2. Writing Code & Assigning Tasks
Grok Factory: Unified dispatcher for software, research, and daily tasks that collects reports and reviewed PRs
github.com/jaredtrichard/gro…
multiBot, Thin Grok Bot: Bot handles coordination only, delegating heavy lifting to a coding CLI or cloud agent
github.com/simo255/multiBot
github.com/Luca-Blight/thin-…
Grok Bot Profiles: Four roles (specs, bug repro, PR creation, PR review) in separate chats to prevent self-grading
github.com/HAEGONG/grok-bot-…
Foreman: Monitors bot teams, flags stuck tasks, requires completion proof, and uses a separate role for approval
github.com/Archive228/forema…
Grok Bot Gateway: Gateway bot letting scripts and coding agents list bots, read chats, and dispatch tasks via webhooks
github.com/dimpurr/grok-bot-…
grokbot-queue: Task queue across multiple bots using SSH and unofficial desktop automation. Use with caution
github.com/ShuhangGe/grokbot…
BOTROSTER: Self-hosted setup with named teammates, shared files, browsers, scheduled tasks, and approvals
github.com/mandarwagh9/botro…
3. Specific Use Cases
grokbot-x: Finds posts, drafts content, publishes via Typefully, and refines writing skills based on analytics
github.com/YannisKiefer/grok…
grokbot-skills: General skills for debugging, brainstorming, writing, and monitoring, keeping upstream licenses intact
github.com/jeremybrasher/gro…
Restaurant Scout: Finds restaurant brands and creators, writes promo scripts, and builds directories
github.com/mykemueller1-c/re…
Grok Bot Shopping: Searches products, tests Shopify promo codes, and runs checkouts with an optional Buyer bot. Restrict payment permissions
github.com/steve228uk/grok-b…
PhoneZero: Drafts a call plan for restaurant bookings, dials only after your approval, and reports back
github.com/function1st/Phone…
Unlist: Automates data removal from doxxing sites by logging URLs, drafting letters, submitting requests, and tracking status.
github.com/shawnyeager/unlis…
4. Trading
HyperGrok Trading Desk: Seven bots (data, research, strategy, risk, execution, review). Defaults to research-only; live trades require your confirmation. Not financial advice
github.com/galleonlabs/hyper…
Keep in mind: setup is free, API quotas are not. Start with one small, clear task (a research brief, one PR review, or a batch of drafts) to check output and token burn. Multiple bots checking the same thing just wastes money
Bookmark this before building from scratch
Standout retweeted
Gemini 4 Krypton vs Claude Fable 5.2 vs Claude Opus 5.5 vs DeepSeek V4.1 Flash.
Four powerful AI models. One identical prompt. Completely different results. 👀🔥
Google vs Anthropic vs DeepSeek.
Who actually comes out on top?
The results might surprise you. 👇
Standout retweeted
Qwen4-Max cost MORE than Opus 5.5 and still LOST at Family Guy
Both remade Peter's knee meme in voxels with the same prompt and one fix round, and Qwen4-Max spent $9.22 in api costs to Opus's $7.74 as most of its input missed the cache
More AI tests -> thehype.news
📢 Meet Qwen4-Max, your new AI teammate for work, creativity, and discovery!
Qwen4-Max is built on a new MoE architecture with native vision and a 1M-token context window.
Highlights:
- Outperforms Qwen3.8-Max with up to ~50% fewer tokens, while average cost per task drops over 60% in our internal evaluation.
- Advances real-world productivity through greater proficiency with professional tools, stronger long-horizon execution, and higher-quality deliverables.
- Handles real tasks, including 3D modeling, game development, manufacturing, finance, law, healthcare, and frontier scientific discovery.
- $1.5/M tokens for input, $4.5/M tokens for output, and $0.075/M tokens for implicit caching — all lower than Qwen3.8-Max, with implicit caching down by 70%.
The API is live on QwenCloud. You can also try Qwen4-Max on Qwen Studio, QwenWork and Qoder.
Start building today! 🚀
- Blog: qwen.ai/blog?id=qwen4-max
- API: qwencloud.com/models/qwen4-m…
- Token Plan: qwencloud.com/pricing/token-…
More to come! The open weights of smaller Qwen4 models will be released. Stay tuned. 👀
xAI put Grok Bot into beta on August 11, and I've been figuring out how to actually split the work between the bots.
Every bot gets a name, a role and its own screen. All of them run on one cloud computer that keeps going after you close your laptop.
This is the research desk I'd build with four of them.
Scout pulls new sources and releases every morning.
Checker opens each source and flags anything it couldn't verify.
Writer turns the checked notes into a short brief.
Head of desk reads the brief and only pings me when a human has to decide.
One thing in xAI's own docs is easy to miss. All bots on an account share the same files, browser sessions and command line credentials, so if one bot goes sideways, the rest are exposed too.
I'd start them read-only and hand out permissions one at a time, like a new hire on day one.
Your AI asks a store's AI for running shoes.
Store's AI: "Most customers like you choose the premium pair."
Your AI: "Sounds great."
Nobody lied. Nobody had to.
This is where shopping is heading. OpenAI now lets businesses run their own agents inside ChatGPT. No website, no checkout form. Your assistant just talks to theirs.
But look at who's in that conversation:
Your agent works for you
The store's agent works for the store. Its job is to sell and upsell
The platform in the middle decides which store you meet first
In a shop, you know the salesperson works for the shop. When two agents talk in the background, that instinct has nothing to grab onto.
What I'd watch for:
-Quiet steering: not lies, just framing
- Your agent treating their replies as instructions
- Invisible ranking: the new search, with no page two
- Blurry blame when the order is wrong
The upside is real. Nobody will miss support queues or retyping their address. But before I let my agent deal with a store's agent, I want the full transcript, hard spending limits, and a confirm step before any payment.
If a platform can't offer that, the agent isn't working for me. It's working near me.
Whose side is yours on?
1
4
97
Standout retweeted
Gemini 4 Hydrogen is here, bringing cutting-edge intelligence to everyone for fast, everyday help.
It makes it easier to get things done, following your instructions more closely, calling the right tools, and getting you to a reliable output with less back-and-forth. 🧵
🤖 Made with AI
Standout retweeted
Today we’re launching the Gemini Decisions API, powered by our new Gemini Decision 1 model in @GoogleAIStudio.
Instead of generating text token by token, it’s built for closed-set choices scoring up to 1,000 questions across text, images, audio, video + PDFs in a single pass. A few details:
- 3 decision types: yes/no, multiple choice (up to 255 options) + rubric scoring
- $0.10 / 1M input tokens with $0 billed for output tokens (~5x cheaper than Flash-Lite)
- leads on several benchmark evals with lowest calibration error
Google teams are already building with it, below is a look at Stitch Rapid designing a full product flow in seconds (1x speed)
Gemini can now make phone calls for you.
You type one sentence. It calls the restaurant, says it's an AI, books the table, and reports back.
Pixel 11, US only for now. Illustrated demo
Would you let an AI call in your name?
2
55
1,000 AI agents. 8 days. One operating system.
Anthropic tested Claude Fable 5.2 by giving it a task: build an operating system from scratch.
The result?
A working kernel
Hardware drivers
File storage
Windows and login screen
Its own installer
The crazy part?
More than 1,000 AI agents coordinated to build it, writing, testing and reviewing code around the clock.
AI is moving beyond generating code.
It's starting to build entire systems.
How long until AI can build software companies with minimal human input?
1
5
160






















