Arian J. Evans

9K posts

Arian J. Evans

@arianevans

Identity/Security Builder: FishNet/WhiteHat Security, Leader: Amazon/Meta. What: (1) Signals Intelligence w/Precision (AI/ML), (2) Low-friction risk mitigation.

A man can't just sit around. Bergabung Ekim 2010

3.8K Mengikuti1.4K Pengikut

Tweet Disematkan

Arian J. Evans@arianevans·18 Oca

Public return (x only) 1/26. Have NOT read DMs, emails, or other social media for ~1-3 years. /Reads automated going forward. If I missed or "ignored" you: not personal. Family, finance, & health came first. Now I swing Build-Hammer hard, again. Smash many rocks. Make diamond.

Arian J. Evans@arianevans

Checking curated X lists weekly now...mainly for AI news. Don't use any other social media and limited email. DMs via Signal. Expect response delays. Don't take misses personally. Laser focused. Family-first.

English

811

Arian J. Evans@arianevans·8h

@essobi Anyone with a parallelized ensemble request router and automated evals (me) knows this definitively. Many interesting ways to confirm. But all you need is basic grasp of arithmetic, econ, & stats 101, to determine probability w/high confidence.

English

Kyle 'esSOBi' Stone@essobi·9h

Are y’all Fscking dumb? Of course they are. So does openAI. So does everyone. These are systems, not bare models. I can prove it, buuuut idk if I care to.

taoki@justalexoki

i 100% guarantee anthropic are routing requests to dumber models at certain times

English

Arian J. Evans@arianevans·8h

@burkov Zuck was/is right on this call - if we measure IG and TikTok influencer impact+ downstream purchases of fake and useless digital MMF and snake oil products. Add AI Impersonation amplification (video, voice): that's our Metaverse.

English

BURKOV@burkov·1d

This screenshot must be shown to every moron repeating AI predictions from billionaires as if they came from some gurus. Just because you were lucky once and it made you rich doesn't make you any different from an average person. You just got lucky once and it made you rich.

Ewan Morrison@MrEwanMorrison

No technology is inevitable or fated. Look at this projection from Zuckerberg from 2022. This week the metaverse got shut down.

English

238

10.5K

Arian J. Evans@arianevans·8h

Will we need Critical human control? At some future date: answer likely "no" for some cohorts. IF Control = TRUE: learning by sliding in dirt = ++optimal human training for 2-wheel emergency survival. Also imminent: Hybrid Human, post-singularity interfaces: e.g.-augmented human vision + transport traction feedback. - Low light and spectrum analysis verify black ice or wet moss in apex; - counter output unlocks brakes, - applies rear torque drift to decrease turn radius This may replace what a skilled rider does reactively by feel and experience today, via inference + from eyes and butt/feet input + human recall.

English

kαι@roguekode·12h

I don’t know how I feel about this. Could be life saving. But at the same time, it seems like it would prevent skill development.

Interesting things@awkwardgoogle

Yamaha self-balancing bike technology.

English

744

Arian J. Evans@arianevans·9h

@HackingLZ Many funded offsec I've seen appear 1-2 years behind & clueless about current SOTA frontier model capabilities & velocity. VC lack AI & Sec depth to evaluate. BUT... smart VC I know bet on "can team spot problems and iterate to success fast?" > over "is current plan viable?"

English

Justin Elze@HackingLZ·1d

I’m really interested in what happens as places take a lot of investment money to build commodity OffSec LLM backed products, even as the barrier to entry keeps dropping. You eventually end up with what actually matters novel research, deep domain expertise, and humans.

English

4.8K

Arian J. Evans@arianevans·9h

@tunguz Is cocaine still popular with city folk? Concerned Ozempic may kill cocaine and jazzercize markets for urban DINK & Corpo Power-Couple cohorts. If true: PLS help identify high-reach, post-Ozempic, Dopamine-amplifier consumesbles.

English

Bojan Tunguz@tunguz·13h

If you want to work on something that is not going to be obsolete soon, look for the inspiration in things that have stood the test of time.

English

2.5K

Arian J. Evans@arianevans·9h

Perhaps I misunderstand your hypothesis -but you can easily confirm or deny dynamic volatility: Do you have a system setup for parallelized n+1 ensemble model request evals? Example eval: 1. Send n+1 Identical requests to model (A) 2. Aggregate & diff (All(n)) responses 3. Request same/any model audit & explain parallel n+1 diff variance. If you want or need any clarification or help just ping me. This is exactly what my underlying abstraction layer (e.g.- for underneath PAI/Fabric) provides for goals like recall enrichment and increasing precision.

English

ᴅᴀɴɪᴇʟ ᴍɪᴇssʟᴇʀ 🛡️@DanielMiessler·11h

Do you have any idea how impossible it would be for Anthropic or OpenAI to constantly change the model they’re serving? They don’t. It’s the same model. When you get different results it’s always for other reasons. - A service interruption - Your harness changed - Etc

dawgyg - WoH@thedawgyg

Opus 4.6 has a bad habit the last few days of outright lying... they must be making it more human... Have had multiple (more than 5) instances where it claims to have completed a task, then when I ask for info about the task and confirmation, it admits t didn't do the task and doesn't have an excuse on why. Then will do the task right then (finally).

English

4.8K

Arian J. Evans@arianevans·1d

@monarch_money Finally: we completely understand the underlying issues that some financial vendors & products are Anti-: aggregator; 3rd party; modern API (/tech), and even anti-functional web design. (Hey Mazda Finance!) Amazed you have so many parts, working well together, already! Thanks,

English

Arian J. Evans@arianevans·1d

@monarch_money But: they ARE recurring. Impossible to complete "review all", or establish 'recurring', from "review recurring". Instead: manually record and track. Locate different UX to create manual 'recurring' entries. Suboptimal. Wound up abandoning. Yet - big customer value in this area!

English

Arian J. Evans@arianevans·1d

@monarch_money is there a way to review recurring merchants without getting blocked by requirement to bind them to an existing account (setup in Monarch)? Is there a way to setup recurring charges without binding them to an institutional account as key value?

English

Arian J. Evans@arianevans·1d

@elkentaro Thank you for your awesome insights into this topic. I will forward your suggestions to my human agents.

English

.@elkentaro·1d

@arianevans @HackingDave you should consider writing an AI assistant to help you

English

Dave Kennedy@HackingDave·1d

I'm getting 10-15 requests a week asking for me to review some innovative new tool for hacking, defense, encryption, home automation - pretty much any topic. I'm getting pull requests that add 40K lines of code with no changes to functionality. Everyone is learning, reminds me so much of the early 90/early 2000 hacking days. I love that everyone is super excited and can now code with their minds, but I do not have the time to go through all of these requests, and I'm moving towards in my open source projects "We do not accept pull requests anymore, give me the claude prompt and desired results and will incorporate and add you to contributors". We live in an interesting/weird time.

English

223

13.4K

Arian J. Evans@arianevans·1d

@DanielMiessler My AE-PTSD 3x Law of Scaling is: For every 3x of raw quant scale: A critical component will melt down or break completely. E.g.- for any baseline 1,000 units: servers, endpoints, requests/sec: =Failure or bottleneck requiring mitigation each 3x of 3x (3000, 9000, 18000, etc).

English

Arian J. Evans@arianevans·1d

@DanielMiessler Suspect Zapier never had to scale before - not like AI+Agents. Everything fails: bowing to the 3x Law. Maintaining request routers, translation layer, and glue code: SUCKS 10/10. I remind myself to leverage existing options, every time I start building one of those tarpits.

English

ᴅᴀɴɪᴇʟ ᴍɪᴇssʟᴇʀ 🛡️@DanielMiessler·2d

Zapier is down. Whenever this happens now I consider building my own version of what that service did for me. And I'm sure I'm not alone. 1. Maybe as good or better 2. One less subscription That's a hard environment to run a service in. You have to be unreproducible.

English

3.5K

Arian J. Evans@arianevans·1d

@tunguz For sure. His question is a tautology, I suspect. He means a specific cohort like "liberal professors of X, at Y". That myopic bet may be sound. IMHO: more likely his cohort / tautological reference becomes irrelevant, via AI impacts you noted.

English

Bojan Tunguz@tunguz·1d

This is already happening as more and more people are doing and publishing their work outside of the usual extremely gate-kept channels. The AI revolution is making it possible.

Bryan Caplan@bryan_caplan

I am open to almost any bet against the humanities and social sciences ceasing to be extremely left-wing anytime in the next 20 years.

English

4.3K

Arian J. Evans@arianevans·1d

@TracketPacer Left SNMP behind ~20 years ago... ...and now close to dirtying myself again for home automation: Generator, power failover, behavioral cascades to devices lacking modern smart/IoT thread/matter type protocols, etc Someone also said +"backup C&C channel". Dangerous blade there.

English

148

TracketPacer@TracketPacer·1d

it’s taken me my entire career up until today to not touch SNMP with my dirty little mitts. i hadn’t so much as blinked at an SNMP trap. but then… i purchased a switched PDU

English

351

11.2K

Arian J. Evans@arianevans·1d

@ZackKorman @bovaird_zach To be fair: every Gen1 admin app/console for everything, uses Impersonation. The only question is: do they later replace it with Authorization, or just hide the Impersonation really really deeply. And claim it's some AuthN or AuthZ complexity whenever issues arise...

English

Zack Korman@ZackKorman·2d

@bovaird_zach I hate that so much. I want ai to do stuff better in its own way not just be a human emulator

English

Zack Korman@ZackKorman·2d

Another AI project making fake security claims. Rescreen claims to "give agents control of your screen, securely." You "decide which apps to give it permissions to" and it "prevents rogue ai agents from doing destructive actions." Using MCP. It doesn't work, obviously.

English

121

6.8K

Arian J. Evans@arianevans·1d

@girdley Yes Am Will Must. Whoa, dino astroids in comments.

English

Michael Girdley@girdley·1d

Temp check: Would you plug an OpenClaw agent into your business TODAY? Why / why not?

English

146

31.3K

Arian J. Evans@arianevans·1d

@_JohnHammond It's not just that AI won't not take your It's Really That will AI job reveal your wasn't doing Nothing value no English major PR, coffee benefits sick day break.

English

John Hammond@_JohnHammond·1d

AI won't job your take but someone who AI your will who take AI usage limits 2x MCP grok slop claude fast skill autoresearch open claw llamas seekdeep agentic million context gpu copilot h̸a̵r̶n̴e̸s̵s̸ b̷̡̲̃̑e̴̜͇͂n̶͉̓̍c̸̳̿h̶̺̗̓̌m̶͇̱̚a̶̙̬̓̕r̶̺̎̈́k̵̻̘̕ w̴̛̘̣͓͋i̸̪͊͂l̴͈̞̤͆̋l̷̥̲͑̐

English

742

54.6K

Jelajahi

@essobi @burkov @HackingLZ @tunguz @monarch_money @elkentaro @HackingDave @DanielMiessler