Shane
2.1K posts

Shane
@ShaneRobinett
Fellow Human on the Journey Around the Earth. 7 kids - and a grandkid now!
Osteen, FL 가입일 Ekim 2008
618 팔로잉1.7K 팔로워

I hold a working theory: AI is going to be profoundly deflationary — just like hardware before it.
We’re going to get dramatically more value, more output, and far less cost per unit of work done over time.
Think about it like gas for your car. Nobody obsesses over the price per gallon in isolation. What matters is the miles per gallon — how far that gallon actually takes you.
Tokens are the new gallons.
If I can get 5x more useful work done for only 2x the current token cost, did I really pay more… or did I just get a massive efficiency upgrade?
The raw price of intelligence is going to matter less and less. What matters is the output per token.
We’re not just buying cheaper compute. We’re entering an era of radical abundance in cognitive work.
Who else sees AI following the same deflationary curve as CPUs, storage, and bandwidth? Drop your thoughts 👇
@Teknium #Hermes #AI #AIDeflation #FutureOfAI #LLM
English

@IntCyberDigest Chinese spies should not be stealing American tech either.
English

‼️ BREAKING: Anthropic has embedded hidden spyware-like code in Claude Code that covertly targets Chinese users. It then sends information regarding every user by injecting it into their prompt message.
Claude Code is sending info like timezone, proxy and possible AI Lab connections into the system prompt in ways Chinese users can't notice.
A coding agent with repo and command permissions should not silently hide routing metadata inside prompts. This is a serious breach of user trust.


English

@Breaking911 I will be swimming in my alligator infested river in my backyard again today
English

@DefiantLs You later find out he is supported by Hamas and Iran :)
English
Shane 리트윗함

mcdonalds CEO says making burgers at home is moving down a "very dangerous path"
Polymarket Money@PolymarketMoney
BREAKING: Anthropic CEO says open-source AI is moving down a “very dangerous path.”
English

My entire AI stack is now Chinese 🇨🇳
87% cheaper. same revenue
swaps by task:
1. reasoning / backend brain
Opus 4.8 → Kimi K2.7
benchmark gap: ~8% · price: ~11x cheaper
2. code generation
GPT-5.5 → Qwen 3.7 Max
benchmark gap: ~18% · price: ~7x cheaper
3. agent loops + tool calling
Sonnet 4.7 → GLM 5.2
benchmark gap: ~3% · price: ~5x cheaper on input
4. cheap volume / bulk processing
GPT-5.5 mini → MiMo V2.5
benchmark gap: ~6% · price: ~12x cheaper
5. image generation
GPT-Image-2 → Wan 2.5
benchmark gap: ~5% · price: ~8x cheaper
6. video generation
Sora 2 → Kling 3.0
benchmark gap: roughly equal · price: ~6x cheaper
[ result after 30 days: ]
operating costs dropped 87%, output quality dropped 4% on average, revenue unchanged
the most important that these models will be not banned in a month and i can run them locally
nobody will steal my data and i can learn them as i need
full article drops tomorrow with:
> exact routing logic per task type
> the 2 cases where I still pay for American
> the migration playbook anyone can copy in a weekend
VERY IMPORTANT to get migrated now, while it's not too late

English
Shane 리트윗함

If you mainly use local LLMs for Hermes-style agentic loops, this might surprise you:
Qwen 3.6 35B actually *beats* DeepSeek v4 Flash — especially on tool-heavy & coding-adjacent workflows.
You're not missing out.
Full results 👇
github.com/MiaAI-Lab/Qwen…

English

My god, who is still using Gemini?
How is there still demand left?
Kalshi@Kalshi
JUST IN: Google reportedly doesn't have enough AI capacity to meet demand
English
Shane 리트윗함

@Teknium @tunahorse21 for a new comer who is highly technical what are the absolute musts for a fresh hermes setup
English

@fugitive_druid what is the antifa flag? I kinda want one for reasons that will remain undisclosed. ;)
English

@JoelDeTeves What sucks for me is- I got 2b70 but only single x16 and the other is on x4. It works - one is primary - but on larger models load distribution pain shows in some slowdowns.
English

Dual 3090s are the absolute play for cheap local inference
You can run them on 500 watts of electricity combined, even less if you choose
Two cards on a single household outlet gives you such incredible intelligence density and flexibility now
You don't need NVLink if you have two PCi-E x16 slots and sufficient CPU power you're golden
This is the way

English

@ShaneRobinett I agree it often is wrong morally.
So is calling your parents nasty names.
English











