@sofiaunchained
Account based inBulgaria
About this account
- Account based in
- Bulgaria
- Connected via
- Bulgaria App Store
Account-level information from X, not a live location or the device used for a specific post.
Host @theaisho | MKT @chainflowPOS & @NishaWeb3AI | Creator @bitcoin_lessons | [email protected]
Europe
Joined January 2012
- Tweets1.4K
- Following1.2K
- Followers479
- Likes3.5K
Sophia ⚡️ retweeted
new paranoia levels unlocked this week:
🔓 bad drones
🔓 supermarket hacker attacks
🔓 https and most of the cryptography we use are cooked (thanks, LLMs)
🔓 hardware wallets sold compromised by official distributor
it's only tuesday ffs
Sophia ⚡️ retweeted
people still don’t get that in the age of super intelligence, your work has to be both great and fast.
i see so many people take 3 days to ship something mid, when AI could’ve made it great in like 10 minutes, and a lot of that is showing up in the AI layoffs.
AI may or may not take your job, but you’re definitely handing it over on a platter.
NEW EPISODE: Models are getting cheaper and more alike, so the real money isn’t in the model itself.
@quxiaoyin broke down why selling “AI skills” doesn’t work, why your prompts and corrections are actually worth something, and the only two business models that seem to be working right now.
The full interview on YT: youtube.com/watch?v=Fz0IuLbX…
. @quxiaoyin was a PM at Meta in 2015, when it was the company everyone wanted to work for.
She later dropped out of Stanford to start Run The World, a virtual events platform that raised $15M from a16z, Founders Fund, GGV and Pear, powered 50,000 events and was acquired by EventMobi.
She says Meta is back, and that Muse could be its biggest win since WhatsApp.
She is now the founder of @agentsky_dev, a marketplace for AI agents. She joined @sofiaunchained and @Diksha_ds on TAS and, and we covered:
- AI models are getting cheaper and more alike, so what still makes money
- Why selling AI skills doesn’t work
- Who will build the best agents two years from now
- Why your AI prompts and corrections are worth money
- Why clean data alone won’t fix enterprise AI
- The only two business models working in AI today
- Why a personal AI assistant has to be free, and who can afford that
- How Muse is cannibalizing Google’s search ads
- Whether training on another lab’s model is wrong
- Why companies won’t pay 100x for similar AI
Tune in…
Sophia ⚡️ retweeted
. @quxiaoyin was a PM at Meta in 2015, when it was the company everyone wanted to work for.
She later dropped out of Stanford to start Run The World, a virtual events platform that raised $15M from a16z, Founders Fund, GGV and Pear, powered 50,000 events and was acquired by EventMobi.
She says Meta is back, and that Muse could be its biggest win since WhatsApp.
She is now the founder of @agentsky_dev, a marketplace for AI agents. She joined @sofiaunchained and @Diksha_ds on TAS and, and we covered:
- AI models are getting cheaper and more alike, so what still makes money
- Why selling AI skills doesn’t work
- Who will build the best agents two years from now
- Why your AI prompts and corrections are worth money
- Why clean data alone won’t fix enterprise AI
- The only two business models working in AI today
- Why a personal AI assistant has to be free, and who can afford that
- How Muse is cannibalizing Google’s search ads
- Whether training on another lab’s model is wrong
- Why companies won’t pay 100x for similar AI
Tune in…
CEOs: AI will change everything in 3 years.
Also CEOs: we funded the model and hoped the team would just figure it out.
JUST IN: 76% of CEOs think AI will change their industry more than anything else in the next three years, according to @Gartner_inc
Gartner says companies often fund the technology and ignore the workforce, which wastes the spend and stalls adoption.
In its example heat map, workslop (low-quality AI output) is the risk that lands in the critical zone and goes to the board.
Full article: gtnr.it/4jrytqv
Sophia ⚡️ retweeted
NEW: Wikimedia Foundation confirmed OpenAI's AI agents made unauthorized test edits to wikis, targeted a citation tool as a possible proxy to pull external data, and tried to compromise its public Etherpad tool.
Massive automated crawling and API requests may have contributed to a May 2026 partial outage of the Wikidata Query Service.
Meanwhile, ships are sinking in our own economic zone. Crew missing. Bulgaria's Prime Minister official line is still “unclear who did it,” and NATO consultations are off the table because the water is "technically neutral". Proud week for Bulgarian science, less proud week for Bulgarian security.
P.S. OSTAVKA, PB!
My proud Bulgarian moment of the week: @OpenAI chose prof. @mvechev, the @INSAITinstitute founder, to join a select group of leading international researchers testing and evaluating a new watermarking technique for AI-generated text.
Sophia ⚡️ retweeted
most AI tools in marketing solve one task at a time e.g. copy, images, and teams end up with disconnected pilots.
the fix starts with mapping the work and the systems it runs on
Is AI making sophisticated investors more sophisticated? 👀
Sat down with @eToro’s Guy Barkat on @theaisho to talk about it.
People hold losing trades for months waiting to break even, and close winning ones in days.
that's from Guy Barkat, who spent 8 years at @eToro going from data analyst to leading API partnerships, watching how 30M+ users actually invest.
Guy came on the show and broke down a few things worth sharing:
- AI is making sophisticated investors more sophisticated. Institutional-grade research used to need an army of analysts, and now one person with the right workflow gets close.
- They launched isolated agentic subaccounts, so you can let an AI manage a slice of your money without touching the rest. You can also connect your account to ChatGPT, Claude or Grok and manage it in plain language.
- And on hiring, Guy says the skill he screens for hasn't changed. judgment. shipping became cheap, so knowing what to ship is the whole job now.
full episode below 👇 w/ @sofiaunchained and @Diksha_ds
worth it if you build fintech products or just want to know where AI investing is actually at
Sophia ⚡️ retweeted
People hold losing trades for months waiting to break even, and close winning ones in days.
that's from Guy Barkat, who spent 8 years at @eToro going from data analyst to leading API partnerships, watching how 30M+ users actually invest.
Guy came on the show and broke down a few things worth sharing:
- AI is making sophisticated investors more sophisticated. Institutional-grade research used to need an army of analysts, and now one person with the right workflow gets close.
- They launched isolated agentic subaccounts, so you can let an AI manage a slice of your money without touching the rest. You can also connect your account to ChatGPT, Claude or Grok and manage it in plain language.
- And on hiring, Guy says the skill he screens for hasn't changed. judgment. shipping became cheap, so knowing what to ship is the whole job now.
full episode below 👇 w/ @sofiaunchained and @Diksha_ds
worth it if you build fintech products or just want to know where AI investing is actually at
Sophia ⚡️ retweeted
The AI Labs Are Worried About Their Own Models xitter.cf/i/broadcasts/1dGYlaoML…
Sophia ⚡️ retweeted
Rabin Nuchtabek runs AI at @hypepartners , a crypto marketing agency, and he's spent over a year in the trenches actually deploying this stuff.
In this chat, @sofiaunchained and @Diksha_ds get into why the model is almost never the real problem, why 95% of AI pilots quietly die, and the genuinely wild setup where his "Grok bot family" books his travel, fills out forms, and trades from its own wallet.
We also get into closed models and lock-in, why he thinks local models are basically "intranet 2.0," the AI slop problem and how he clones his own writing voice, and his take on why the next high-paying AI job won't even show up as a job posting.
full episode 👇
Sophia ⚡️ retweeted
BlockZero returns to London on November 15, part of @Solana Breakpoint community events. Details below👇
BlockZero 2026 is coming to London on Nov 15.
A day for the people running the infrastructure: validators, operators, protocol contributors, and the teams building validator-adjacent tech.
Small by design, attendance is capped and subject to approval.
luma.com/dmxdc394
Sophia ⚡️ retweeted
Most LSTs stack fees: the pool takes one, the validator takes another.
chainSOL doesn't. We return 100% of our commission and MEV to holders, every epoch, on-chain.
Sophia ⚡️ retweeted
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must…
The weekly sit-down with @Diksha_ds discussing the latest stuff on the AI front
"AI Could Kill Everyone" + OpenAI's Math Flex and the Money Getting Insane xitter.cf/i/broadcasts/1aKbdEYor…
Happy Monday with some AI news.
"AI Could Kill Everyone" + OpenAI's Math Flex and the Money Getting Insane xitter.cf/i/broadcasts/1aKbdEYor…
Sophia ⚡️ retweeted
Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.
The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible - but I hear the same people express fear privately. No other human activity poses this level of danger.
