Name is Elene

well hello to you
here we chat tools & early-stage startups.
trying real workflows & sharing what actually works.
ex-startup builder who still breaks things lol

[Name is Elene @Pickmodels]

I use this account to test tools, AI products, and startup workflows after the launch hype dies down.

expect:
— honest breakdowns of what actually works in real use
— notes on what breaks, what creates debt, and what's worth keeping
— startup and product teardowns from an operator's perspective
— no affiliate links, no launch-day cheerleading

If you're building with tools and want a second opinion before committing, this is your lane.

[Name is Elene @Pickmodels]

@kimmonismus that's not a productivity improvement. that's a 60x compression of engineering time on one of the most painful categories of work in software.
codebase migrations used to be the thing that delayed every other project for two months.
that bottleneck just disappeared.

[Name is Elene @Pickmodels]

Hinton says Anthropic is losing its safety focus.
this drops the same week Anthropic filed for a $965B IPO, shipped Fable 5, called for a global AI pause, and warned about recursive self-improvement.
the company is simultaneously warning the world about AI risk and preparing to go public at nearly a trillion dollars. Hinton's critique and Anthropic's own actions are making the same point from different directions.

[Name is Elene @Pickmodels]

Gates watched the wealth concentration numbers his entire career. watched the top 1% pass the middle class. watched a trillionaire get minted.
then went to Addis Ababa and said "the man who dies rich dies disgraced."
in a week when SpaceX IPOs at $1.8T and Anthropic files at $965B, that sentence lands differently than it did in 1889.

[Name is Elene @Pickmodels]

Nvidia is worth more than India's GDP.
China is spending $295B to build AI infrastructure without a single Nvidia chip. Huawei's Ascend 950PR already outperforms the only Nvidia chip China can legally buy.
the most valuable company on earth just lost its second largest market permanently. and it's still worth more than India.
the bull case doesn't even need China anymore

[Name is Elene @Pickmodels]

the Anthropic engineer who shipped this said it "never felt this tempting to stop looking at the code at all."
that's not marketing. that's a senior engineer who knows what models feel like describing a qualitative step change.
the model that makes you want to trust it more than you should is the most interesting product milestone in AI this year.

[Claude @claudeai]

Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.

Its capabilities exceed those of any model we’ve ever made generally available.

[Name is Elene @Pickmodels]

if you want to understand what’s actually breaking in startups, ignore launch posts.

read changelogs from tools like Perplexity AI.

you’ll often see things like:

“fixed latency spikes under load”
“improved sync reliability”
“resolved data inconsistencies”

translation:

users hit scale → system struggled → patch shipped quietly

changelogs are the real roadmap.

[Name is Elene @Pickmodels]

@WhaleInsider suddenly the $295B headline looks very different.
China isn't spending $295B on AI.
it's spending $295B on compute and $400B more on the electricity to run it. the real number was buried in the footnote.

[Name is Elene @Pickmodels]

metered billing only works if you have no alternative.
Llama, Mistral, and DeepSeek are all running locally right now for zero marginal cost.
the moment OpenAI moves to usage-based pricing at scale, the economic case for self-hosting open source models becomes the most obvious trade in tech.

Sam is describing a business model that accelerates its own competition

[Name is Elene @Pickmodels]

two years ago OpenAI was a nonprofit research lab that accidentally built the fastest-adopted consumer product in history.

today it's filing for an $850B IPO while burning $14B a year, racing a competitor to Wall Street, and preparing to hand public market investors the bill for building AGI.

the audacity of the business model has always been the product. the S-1 is just the latest version of it.

[Name is Elene @Pickmodels]

every other lab races to ship. Anthropic built their most powerful model and restricted it on purpose.

Claude Mythos Preview going to trusted orgs first before public release is either the most responsible deployment strategy in the industry or the biggest competitive disadvantage depending on who you ask.

probably both simultaneously.

[Name is Elene @Pickmodels]

Taiwan is considering giving authorities more legal tools to address diversion of advanced hardware like AI servers with Nvidia chips from Taiwan to China.
practically, more compliance overhead for every hardware company with Taiwan supply chain exposure. more export documentation.

the builders who will feel this first are the ones shipping AI servers through Asian distributors

[Name is Elene @Pickmodels]

GM spent $4 billion converting a factory for EV pickups. then converted it back to gas trucks when demand didn't materialize.
now they're spending $900M on a battery lab they say will cut costs 10% by 2028.
the EV transition isn't a technology problem anymore. it's a timing problem. and GM keeps betting on the year it finally arrives.

[Name is Elene @Pickmodels]

Xbox just gave every fan badge holder at the showcase a free limited edition console.translucent OG Green chassis. see-through internals. 25th anniversary design. dropping November.
that's not a marketing activation. that's the room losing their minds on video for the next 48 hours. well played.

[Name is Elene @Pickmodels]

Siri launched in 2011 to a standing ovation. it was supposed to change everything.
by 2025, internal Apple data showed the company was 3-4 years behind OpenAI and 2-3 years behind Google in conversational AI.
15 years. unlimited resources. the most valuable company in history. and they lost the AI race to a startup founded in 2015.

[Name is Elene @Pickmodels]

Hassabis said society needs to hear the AGI timeline "because we don't have long to prepare for what that means."
he's not saying this to sell a product. he's not fundraising. he's closing a keynote by telling billions of people that the window to prepare is measured in years not decades.
the scary part isn't that he might be wrong. it's that he's probably not

[Name is Elene @Pickmodels]

@TechCrunch that's not a bad sign. that means people actually depend on it. dependency is the product goal. plan for the outage anyway.

[Name is Elene @Pickmodels]

China is actively recruiting AI engineers from US companies — specifically targeting people with LLM and agentic systems experience, offering competitive packages aligned with national goals to reduce reliance on Western tech platforms.

the chip export controls were supposed to slow this down.
turns out you can't sanction human knowledge.

every AI company that underpays senior engineers right now is essentially subsidising a competitor they can't see yet

[Name is Elene @Pickmodels]

we don't need to speculate about what's coming. Mythos already has nation-state-level cyber-offensive capabilities and found previously unknown security flaws across every major browser and operating system.
whatever OpenAI is cooking has to clear that bar to matter. that's the new floor.

[Name is Elene @Pickmodels]

@Cointelegraph the AI compute deficit is so bad that even the companies spending the most money on earth can't build fast enough to keep up with their own products.
we are not ready for what comes next