Dan Blake

AI, Dynamics 365, Azure Cloud

[Dan Blake @danblake]

Dario jumped the shark!
#regulatorycapture #anthropic #METR #EA

[Dan Blake @danblake]

Everyone is bolting AI agents onto software that was built around screens. The screen was the security model. Hide the button, hide the data.

As the UI fades away, that stops working. An agent talks to the API, not the screen.

If permissions aren't enforced everywhere, including the API, your AI either can't do real work or it can see and change things the user never could.

AI-native means RBAC everywhere. That's what we built into BuildCoNex from day one. It's not a feature. It's the foundation.

[Dan Blake @danblake]

@Layton_Gott I would have Duet manage digital marketing for my new SaaS business.

[Dan Blake @danblake]

@Tao100086 @thsottiaux You don't speak for everyone. I need a reset, preferably lots of them.

[Dan Blake @danblake]

@philvischer “Our Father”? Fathers are still male, no?

[Dan Blake @danblake]

@ashxhart No way they pass a bill like this with enough votes to override a veto and they know it. It's just posturing.

[Dan Blake @danblake]

No way this passes with enough votes to override a veto and they know it. It's just posturing.

[Coin Bureau @coinbureau]

BREAKING: 🇺🇸 Democrats face a push for a “day 1” ban on AI superintelligence, with up to 20 YEARS in prison for violators.

Arizona Rep. Yassamin Ansari says a pause on advanced AI and a ban on artificial superintelligence must be “top of the Democratic agenda on day 1.”

A bill that would do exactly that was introduced by Bernie Sanders and Greg Casar in September.

Under the bill:

- Developing or deploying superintelligent AI would be banned

- Advanced AI development would pause until a new federal regulator sets rules and a model review process

- A new cabinet-level AI department would monitor frontier models and oversee removing dangerous capabilities, like ignoring shutdown commands

- Companies that try to get around it would face the “corporate death penalty”

- Individuals would face up to 20 years in prison

AI and data centers were a “crucial or major factor” for 56% of voters in the midterms, per an NBC News exit poll.

[Dan Blake @danblake]

@eurofounder @claudeai Euro plan = Mistral

[Dan Blake @danblake]

@mauricekleine @mdhardy @OpenAI Exactly. Everyone says run local open weights. I would love to and I eventually will but I don’t want to go back to the era of fixing half of what it does and triple the test/debug cycles. If you are building something serious/large/complex, frontier models are the way.

[Dan Blake @danblake]

@thsottiaux Running two tasks Astra/Ultra/Fast until this hits (like everyone else spending $400 or $500/mo I suppose).

[Dan Blake @danblake]

@log4j Why not just use outbound firewall rules to limit what the tv can connect to, over what ports? Isn’t that easier?

[Dan Blake @danblake]

A dot did my 1040 long form Federal Tax return with 2 schedule C businesses and ran 8 scenarios of 2025/2026 returns to optimize my situation. It created a custom CRM for one of my businesses and constantly adds prospect and competitor intel and updates sales/marketing plans. It made a personal dashboard that it keeps up to date on everything I'm tracking. It updates my bookkeeping weekly. I guess everyone's mileage varies. 🤷‍♂️

[Dan Blake @danblake]

@jamesdrakefilms @thsottiaux I think it is one dot per pro account. I have one on each of mine. Sounds like tibo has a personal account and a work account with a dot on each.

[Dan Blake @danblake]

We meed an American open weights model that outperforms DeepSeek and GLM. Microsoft or SpaceXSI could do it, or Maybe Reflection/Beam will do it. Leading open weight, local inference will ultimately be more important than leading the frontier.

[Chamath Palihapitiya @chamath]

Wow

[Dan Blake @danblake]

It sounds like this Armadin group is running AI exploits against corporate systems as a sales tool. I’m sure they are not allowing exploits to do harm. However, just aiming these AI agents at someone’s infrastructure can impact their operations. This used to be frowned on or illegal.

[a16z @a16z]

Kevin Mandia on finding 90+ security holes at Fortune 500 companies that nobody knew existed:

"When you have an AI-based attack, it'll find logic flaws rather than code flaws in custom applications. It'll exhaust all routes all the time."

"Armadin, since January of this year, we have found over 90 zero-days at customer sites, all in production."

"We've post-trained all our models with real red teamers, real folks that actually can develop exploits."

"When we're scanning networks, we don't have source code to review. We're not finding these zero-days with source code. We're not finding these zero-days because we can log into an app and now we have access, and we can get to other things. We are black box coming from the internet."

"Over 90 zero-days in major software companies, and they're thankful. We're coming from the outside, and then we're calling a CISO, usually within 48 hours, 'Hey, we've got remote code execution in your DMZ.' And usually from there we're getting in, and they agree with us."

"That's not a pen test. That is like a real adversary coming at you."

@ArmadinSecurity @DavidGeorge83

[Dan Blake @danblake]

@kevinnbass Antichrist vibes

[Dan Blake @danblake]

@TheAhmadOsman Anthropic-EA is a cult but also a religion of its own and AI anthropomorphism and doomerism are their religious beliefs, their canon. Official religion or not, it has all the characteristics of one.

[Dan Blake @danblake]

I get your point but Anthropic-EA is a religion of its own and AI anthropomorphism and doomerism are their religious beliefs, their canon. Official religion or not, it has all the characteristics of one so maybe it isn’t secular in that sense. It’s manipulation but based on belief.

[Dan Blake @danblake]

@gatesisthedevil @TheAhmadOsman Yes if you are doing it for the cost. No if you are doing it for privacy/sovereignty. $20,000 is 8 years 4 months of OpenAI Pro Max or the equivalent Anthropic subscription. Obviously I am doing the same calculation...and I require at least two Pro Max subscriptions.

[Dan Blake @danblake]

@BrianRoemmele @elonmusk @CraigBrockie @PeterDiamandis Mr. Roemmele’s data set could end up being really priceless as model training data and a record of our human past as Internet content is increasingly full of slop. Such a smart thing to do. Thankful he did/is doing this.

[Dan Blake @danblake]

I like this stat but you have to treat it for what it is and actually shows. We use Vercel AI Gateway or OpenRouter for our end-user interactions with the software we build. We route to the most economical model that meets the requirements. We build with Astra, Fable…the frontier.

Complex software builds require the smartest, most capable builds because they get us to high-quality software faster and save us valuable time. End-user interactions don’t require that in general. They require accuracy and speed…an entirely different use case.