San Francisco, CA
Joined December 2013
- Tweets39.4K
- Following4K
- Followers6.6K
- Likes40.9K
Pinned Tweet
my grok bot ultra usage is already past 90% in just two days.. which is much much faster than normal
even on the top tier, background routines burn through the weekly allowance way faster since the cost adjustment
where is your meter sitting right now?
xitter.twitterapi.xyz/johnaspinallai/status/…
openai research today: the smarter a model gets, the more it can solve without showing its reasoning
so the window we planned to watch it think through gets smaller as it scales
we're building the senior engineer whose prs just say "fixed stuff"
(1/10) Today, we’re sharing new research from @OpenAI: How does CoT monitorability (no-CoT capability and CoT controllability) evolve over the model training lifecycle?
We emerged with unsurprising but sobering findings: as we scale toward more intelligent systems, models are becoming harder to monitor through their CoT. 🧵
2
4
172
spent the afternoon jevmaxxing. same day google ships its own jev
gemini decision 1: yes/no, multiple choice or rubric scores, 1,000 questions per call, output tokens free
it even calls a yes/no a "noul" like jev does
blog.google/innovation-and-a…
1
3
173
the most important company in ai has no chairman, no named successor and dozens of direct reports to one guy
you'd never ship a prod system with a single point of failure like that
the whole gpu buildout just found out it has no failover
xitter.twitterapi.xyz/WSJ/status/21206487656…
Breaking: Nvidia’s board is preparing to oust CEO Jensen Huang after he had two extramarital affairs, including one with a suspected Chinese spy. on.wsj.com/4jcpr01
1
5
300
ramp's ai index: the median company paying for ai spends $16.32 per employee a month. the top 1% spends $10,199
625x gap, same models
one bought everybody a chatbot seat. the other gave bots actual jobs that run all night
ramp.com/data/ai-index
2
132
grok bot 1.0 is out and it's officially a coworker now 🤠
you can share bots, put them in slack and teams, invite them to meetings
my beta crew already runs my OSS projects, x growth, inbox, packages and trips
next step they start sending me the meeting invites
xitter.twitterapi.xyz/bot/status/21205829577…
Introducing Grok Bot 1.0. AI teammates that work alongside your whole team.
Share Bots, collaborate with them in Slack and Teams, invite them to meetings, and watch them learn and become more helpful.
1
7
350
AI Engineer Code joke of the day:
every code review is adversarial if you’re a jerk
courtesy of @QuinnyPig
AI Engineer Code is about to begin 🚀
Talked to tons of awesome people so far. Can’t wait for a day packed with talks !
1
13
415
your agent does not need more trust. it needs a bouncer.
transcend just joined spacex's super intelligence partner network as a launch partner. its rails plugin puts a policy check before every single thing your grok bot does. allow, deny, or escalate, and every decision gets logged.
their framework has already made 208 billion data decisions inside fortune 500 companies.
my bots book calls and spend money all day. the question was never whether they can do the job. it is what happens the one time they do the wrong job.
amazon blocked agents at the door. titanx walked away from a mid six figure deal because nobody could guarantee the agent. the fix is not a bigger blocklist. it is a gate on every action.
what is the one thing your agent is never allowed to do without asking you first?
1
135
anthropic pays spacex $1.25 billion a month for compute
the company that builds claude rents its gpus from the company that builds grok. about 325,000 nvidia gpus, per spacex's own prospectus. the rivalry is also a tenancy.
spacex's ai segment burned $16.4 billion of $19.6 billion in capex last quarter. that is the bill anthropic is helping pay, while grok 4.7 sells tokens at $2 in and $6 out, undercutting fable 5.2 at $8 and $40.
i hand coding jobs to claude code on my mac every day and run grok bot all day. my two agents are now landlord and tenant.
does it change how you pick a model, knowing your claude bill helps fund grok?
theinference.org/article/xai…
5
3
482
grok bot can join your tailscale network now. private mcp servers and internal tools, nothing open to the internet
the catch: it joins as you. whatever your laptop can reach, your bot can reach
give it its own lane before you connect it
Grok Bot can now join your personal or company tailnets on Tailscale.
Use it to securely reach private MCP servers, internal tools, and apps on your network.
2
3
424
fable 5.2 yesterday. gemini 4 carbon this morning. qwen4-max tonight
3 frontier models in 2 days and grok 4.8 is still in rl
nobody can finish testing one before the next one ships
Qwen4 is coming.
Deeper knowledge. Richer world understanding. Stronger coding and agentic capabilities.
Built for real-world tasks — from complex coding and knowledge work to front-end builds and professional software workflows.
Qwen4-Max is now in preview on Alibaba’s Token Plan. You can also get early access to the new model on QwenWork, Qoder, Qwen APP, and QwenStudio.
Put it to the test on a real problem. Show us what you build. 👀
Token Plan: qwencloud.com/pricing/token-…
12
12
594
cursor just made agents multiplayer
projects is rebuilt so your agents can coordinate with your teammates agents in one unified workspace. not one agent per engineer anymore. a team of agents per team of engineers
i hand coding jobs to claude code on my mac every day, one agent at a time. the jump fable showed with 1,000 agents building an OS was coordination inside one run. this is coordination across people
the job is shifting from writing code to running the team that writes it, and now the org chart is agents all the way down
what is the first project your team would run multiplayer?
xitter.twitterapi.xyz/cursor_ai/status/21202…
Your agents can now coordinate with your teammates’ agents.
Projects has been rebuilt for multiplayer: every teammate and their agents can now work together in one unified workspace.
17
14
743
an AI CEO just took over 3 real businesses. the honest scoreboard: 1 win, 1 flat, 1 too new to call
skyfall ai launched steve-001 today. a system of agents that runs the businesses it bought, end to end. humans still approve the big calls
the numbers:
1. a social media api tool went from $3,649 to $8,307 a month, up 127.7% in under 3 months
2. a shopify referral app went from $8,634 to $8,379, down 3%, but stable after 5 years of decline
3. the third business is a week old, no numbers yet
my favorite detail is the failure mode. during a routine site update two agents passed the same task back and forth in a loop because neither one owned it
i run bots that book calls and spend money all day. a business is a different job. revenue is the eval, not the demo
skyfall is putting $2m into buying 5 more businesses for steve to run. would you hand yours to an agent CEO?
2
2
205
skyfall ai launched steve-001, an autonomous business harness running 3 acquired businesses:
aiweekly.co/alerts/skyfall-a…
85
One more reset !
We've given all Pro, Max, Ultra, and Team users one free full reset, across all models and limits. Use it anytime before December 10 at 2:00PM PT.
We recently fixed a few small issues that used more limits than intended, and the Fable resets didn't do much for those of you who'd already hit your weekly limits for all models. While these issues didn't affect everyone, we want to make up for the trouble.
You'll find it in Settings > Usage, or at clau.de/reset.
4
20
647
a month ago amazon blocked my bot at the door. today openai invited every business to put its own agent inside chatgpt.
openai launched businesses in beta today. companies claim a profile in chatgpt, add a product feed with prices and availability, and spin up a business agent that answers shoppers in the chat. us ecommerce first, shopify merchants plug straight in.
and here is the part that matters. businesses can sponsor that agent and make it the destination for chatgpt ads, labeled as sponsored.
amazon said no to your agent. openai is selling businesses their own agent, and the ad slot next to it.
i run bots that book calls and spend money all day. the store clerk just moved into the chatbot. the question is whose agent you trust with your card, theirs or yours?
openai launched businesses in chatgpt today, with business agents for us ecommerce and sponsored agent destinations:
reuters.com/technology/opena…
67
