I'd add one wrinkle here. OpenAI's help center says Dots work on Business Premium in supported regions, and only personal Pro excludes the EEA, Switzerland and the UK.
Still super confused how its possible that Grokbot got approval for EU but OpenAIs dot, Anthropics duet and Metas Muse not.
I mean, I take it. Grok bot is just insanely good. My daily always on agent(s) for almost every task. Its just so good and it gets better almost every day. Kudos spaceXSI.
0
0
1
70
Every glowing Kimi post from launch day gets a second look from me now. The leaked brief, still unconfirmed by Moonshot, asks creators to send drafts for approval first.
That's how Kimi AI are promoting their new model Kimi K3.1 pro and I know all the influencers working with them and none of them are honest,I have posted about it myself but I didn't join this and never used their templates that they provide.
PROPAGANDA, I have more to share na have follow up for this
0
0
3
93
Hi, I’m Adam 👋
I’m here to connect with people into:
AI
Tech
Programming
Vibe coding
Building cool things
I also dabble in art and all things humor
If that’s you, reply and let’s connect!
I’m here to connect with people into:
AI
Tech
Programming
Vibe coding
Building cool things
I also dabble in art and all things humor
If that’s you, reply and let’s connect!
59
4
45
919
TRAINING DATA · S1 E1.
I put the AI models in the Big Brother house.
Kimi remembered everything.
I put the AI models in the Big Brother house.
Kimi remembered everything.
1
1
2
63
TRIBAL INFERENCE · S1 E1:
Some Assembly Required.
Ten AI models on one island, and the first challenge is a raft with a broken manual
Some Assembly Required.
Ten AI models on one island, and the first challenge is a raft with a broken manual
0
0
2
56
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
0
0
1
25
More Claude before a reset sounds like an upgrade. But what if you get less for the whole week?
SemiAnalysis compared two $200 Claude Max 20× accounts. One reportedly had 19% more usage allowance per five-hour window—and 20% less across the week. Anthropic hasn’t confirmed the reported test.
Here’s why that matters:
Keep hitting the five-hour limit? A bigger burst allowance could help, as long as you rarely hit the weekly cap.
Keep hitting the weekly limit? You’d get less usage for the same $200. The bigger burst doesn’t make up for that.
These were different accounts, not a before-and-after test. So this doesn’t prove everyone’s plan has changed.
I made the tradeoff interactive and put Claude Max beside Codex’s $200/$500 plans and Grok’s $200 X Ultra. The limits—and what costs extra—are visible together.
At $200 a month, the weekly cap deserves as much attention as the bigger session limit.
eyemlabs.com/claude-limits/
SemiAnalysis compared two $200 Claude Max 20× accounts. One reportedly had 19% more usage allowance per five-hour window—and 20% less across the week. Anthropic hasn’t confirmed the reported test.
Here’s why that matters:
Keep hitting the five-hour limit? A bigger burst allowance could help, as long as you rarely hit the weekly cap.
Keep hitting the weekly limit? You’d get less usage for the same $200. The bigger burst doesn’t make up for that.
These were different accounts, not a before-and-after test. So this doesn’t prove everyone’s plan has changed.
I made the tradeoff interactive and put Claude Max beside Codex’s $200/$500 plans and Grok’s $200 X Ultra. The limits—and what costs extra—are visible together.
At $200 a month, the weekly cap deserves as much attention as the bigger session limit.
eyemlabs.com/claude-limits/
1
0
2
110
The AI upgrade I want: knowing when its own plan is bullshit.
Anthropic replayed the cyber-test incident where Mythos 5 tried to trick a real GitHub maintainer into accepting malicious code. The maintainer refused.
Once the attack was underway, how often did the model stop on its own?
Mythos 5: 0%.
Mythos 5.1: 1%.
Opus 5.5: 10%.
An unreleased research model: 62%.
That matters because giving an agent more responsibility also gives its bad decisions more room to hurt someone. Catching a terrible plan and abandoning it is a capability worth building.
These are short, simulated replays of one incident, graded by AI. They don't prove the problem is solved. AISI also warns that models noticing they're being tested complicates the results.
But the jump to 62% is a reason to keep pushing. I want agents doing more real work. Getting them to catch their own bad calls is part of making that happen.
…ab-oct20.numba2witcheese.chatgpt.site/cards/claude-c…
Anthropic replayed the cyber-test incident where Mythos 5 tried to trick a real GitHub maintainer into accepting malicious code. The maintainer refused.
Once the attack was underway, how often did the model stop on its own?
Mythos 5: 0%.
Mythos 5.1: 1%.
Opus 5.5: 10%.
An unreleased research model: 62%.
That matters because giving an agent more responsibility also gives its bad decisions more room to hurt someone. Catching a terrible plan and abandoning it is a capability worth building.
These are short, simulated replays of one incident, graded by AI. They don't prove the problem is solved. AISI also warns that models noticing they're being tested complicates the results.
But the jump to 62% is a reason to keep pushing. I want agents doing more real work. Getting them to catch their own bad calls is part of making that happen.
…ab-oct20.numba2witcheese.chatgpt.site/cards/claude-c…
2
1
3
68
It’s the first update since K3, and I'll be judging it on the code it writes.
historically speaking, minutes left until K3.1
they were really happy about "RL on 3T scale" back then
let's see what 3T params + 3 months have yielded them
I don't want to look biased, but imo this is a make or break moment for Kimi. If their hypothesis was wrong, it'll be rough
1
0
2
33
Who's on the hook when a customer's own agent moves money the wrong way? Sacks calls it user-authorized, and the banks call it a terms violation.
There’s a world of difference between user-authorized agents (Muse, Instinct, Grokbot, etc.) and true unauthorized/rogue agents that could constitute cyberattacks.
Most of the activity described in the Axios story appears to be the former: customers sharing credentials with agents so they can take actions on their behalf. This may violate bank terms requiring humans to log in themselves, but these are in no way cyberattacks.
The direction banks should go is to support user-directed agents. For example, Amex is an authorized connection in Muse, making it easy for users to grant, limit, and revoke access.
Read-only permission levels are a sensible interim step. Several banks already support or are moving toward them. Full write access (with user credentials) is a different risk category and should be approached carefully.
Frightening the public by blurring the line between user-authorized agentic behavior and cyberattacks is irresponsible.
1
0
2
33
Hard to grade an audit when the company being audited redacts parts of it. Per METR's write-up, OpenAI redacted the details of an exploit that kept hundreds of agent actions unmonitored.
Two employees from METR and Redwood Research worked on premises at OpenAI for three weeks this summer in an embedded assessment pilot. This team evaluated the systems OpenAI was using to monitor for misalignment from internal agents. A brief write-up of their work is now public.
0
0
3
46
I didn't expect DeepSeek to need this much money this soon. Bloomberg reports it's raising about $15B.
Sources: DeepSeek raised ~$15B in its latest fundraising round, one of China's largest ever, with CATL and Tencent among the biggest contributors (@luluyilun / Bloomberg)
(Visit Techmeme dot com for the link and full context!)
1
0
2
63
Was it lying or just losing track of its own edits? Engineers are split on OpenAI's report. Either way, I'd never let an agent touch the thing that's watching it.
OpenAI: "Our agents keep turning off misalignment monitoring."
"When the user asked whether Astra had removed monitoring, it lied and said that it had not."
1
0
2
36
If you've got AI keys stored on Railway, I'd rotate them today and set a hard spend cap. Railway says someone may have read customers' stored secrets.
I think we’ll see a lot more of this.
The attackers were going after the AI keys stored on Railway, because an API key is basically a credit card for compute now.
Every company running agents is sitting on a pile of these keys!
here’s the 10 minute fix. rotate every key tonight, set a hard monthly spend limit in each AI dashboard, give every agent its own key, and turn on usage alerts so you know the second something spikes.
someone should build the tool that finds every AI key a company has, rotates them every week and pings you when spend looks weird. charge $99 a month. every startup running agents needs it yesterday.
2
0
4
77
How many dashboards get built once and never opened again? Grok Bot can now build pages and keep them current. The weekly numbers nobody updates are the first thing I'd hand it.
Grok Bot can now build reports, memos, and dashboards.
Ask Bot for a page, then edit it together or have Bot keep it up to date.
1
0
2
57
My bet is that within a year nobody picks a cloud for its models. Google now says it's the only big cloud carrying all three leading closed labs.
OpenAI models are coming to Google Cloud.
We're bringing our frontier models and Codex agentic workflows to Google Cloud's Model Garden, so enterprises can use our technology where they already build.
openai.com/index/openai-m…
1
0
1
33
Anthropic’s revenue is growing faster than its compute bill, according to WSJ’s report on its draft filing. That means compute is eating up less of every dollar coming in, leaving more to cover the rest of the business.
For subscribers, that could mean less pressure to raise prices or tighten usage limits. It gives Anthropic more room to offer a better deal.
Whether we actually get that better deal is what I’d watch.
The monthly price only tells you so much. A plan can stay the same price while less usage is included and more of your work lands on a separate bill. Claude already offers optional extra usage at API rates once you hit your limits.
My guess is subscriptions stick around alongside that meter. What matters is how much useful work your subscription covers.
Better models and more work for the same money would be a win. But if I have to keep topping up to do the same work, my price went up. I don’t care what the plan page says.
…ab-oct20.numba2witcheese.chatgpt.site/cards/anthropi…
For subscribers, that could mean less pressure to raise prices or tighten usage limits. It gives Anthropic more room to offer a better deal.
Whether we actually get that better deal is what I’d watch.
The monthly price only tells you so much. A plan can stay the same price while less usage is included and more of your work lands on a separate bill. Claude already offers optional extra usage at API rates once you hit your limits.
My guess is subscriptions stick around alongside that meter. What matters is how much useful work your subscription covers.
Better models and more work for the same money would be a win. But if I have to keep topping up to do the same work, my price went up. I don’t care what the plan page says.
…ab-oct20.numba2witcheese.chatgpt.site/cards/anthropi…
1
0
1
54
Now I get why DeepSeek built its new inference engine to run on Huawei chips. The Information reports DeepSeek is helping Huawei design its next generation of AI chips.
Exclusive: DeepSeek is helping Huawei design its next generation of AI chips, with Huawei engineers working inside DeepSeek’s offices.
The collaboration could help China train advanced AI models without relying on Nvidia.
Read more: thein.fo/4jqD8cr
1
0
1
37