Pandah (🥃,🦉)🐸🧙🏽‍♂️

Gaming/Web3 is my passion. Here to feed my addiction of researching, reading and listening to podcasts/spaces daily.

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

Everyone's chasing immediate monetary gains. Missing the forest for the trees.

[VincentVanDough @Vince_Van_Dough]

The best art is made for the sake of making art, not making money. The money side of things will take care of itself in time if the work is good. The market shouldn’t dictate your output.

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

this is actually insane, most people are taking peptides without any oversight or trying to eat healthy to prevent any future diseases.

if you had your whole genome sequenced and read with ai you could actually hack yourself to a better healthier future instead of taking guesses

wild what's happening in the biotech field as most are focused on only tech

[The Genome Computer Company @genomecomputer]

Today we’re introducing .genome, the first genome data file designed to be read by AI.

.genome requires 3-10x fewer tokens than VCF and improves factual accuracy by 10-20x, meaning AI can answer faster, cheaper and more accurately.

Together with .genome we’re releasing READMYGENOME.md, a Claude skill file built specifically to help AI read and interpret genomes.

Both are open-source, Apache-licensed, and at https://t.co/gpCveGcQ4s.

If you don’t have your genome sequenced, visit https://t.co/8aBNuC9svO and order a kit.

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

pump stocks > tokenize stocks on chain > use crypto bros as exit liq

maybe just maybe

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

@huuep insane

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

spintronics is the alien tech everyone's waiting for

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

every human is an api
your relationship is the end point.

good apis cost more than cheap ones to call.
the cost is not calculated in money but humanity.

trust is the api authentication layer.
friendship/relationships are the subscription.
vibes are the inference/bandwidth/rate limits.

valuing everything is money is the trap.
money isn't the enemy but its a terrible master.
it's cheap abundant and man made.

your mind was not meant to calculate or think in a framework governed by money as your sole OS.
break free

-braindump

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

@0xSero Congrats

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

web3 is flooded with countless speculative plays that the community gambles on. but here's someone actually delivering:

@0xSero is making frontier level moe models (like glm-4.7) runnable on consumer hardware through serious one shot expert pruning (reap) + quantization + smart runtime swapping.

donate with either your brainpower (test/contribute to https://t.co/EIGCi4o8E4), funds or just spread the word.

we need way more moe pruning, quantization innovation, and local first tooling to get smarter models in everyone's hands not locked behind apis (which are bound to get more expensive as energy and demand grows). Spend your spare time deep diving local llms+ hardware instead of watching endless ai slop content.

ribbit! 🐸

[0xSero @0xSero]

https://t.co/txDnDa0Flf

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

@xyster @LottoLabs Hey Steve could you send me a dm as well.

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

@NickPlaysCrypto @avax Woof? lol

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

@udiWertheimer they can have 3 mac studios with over 1tb of unified memory running deepseek locally. would save soo much cost!

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

@farokh death by dorito

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

@farokh big spreadsheets guy

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

with all this bot slop on the interweb now... will they prop up proof of human as the next meta?

maybe just maybe that will require all humans to KYC/proof of human

maybe that was the plan all along i think it all comes full circle. aren't you sick of all these bots? don't you want a solution? just give us your fingerprints, eyeballs and a mugshot.

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

ai is moving much like anything else that starts off as an everything deal and slowly creeps into more niche substrates

General intelligence > specialized intelligence > niche intelligence

the next big thing in the headlines will be edge devices IMO powered by specialized llms that cater to a specific task/job

you wont need server grade hardware to solve specific problems autonomously, even things like raspberry pi's, arduino or even specifically manufactured integrated circuits/IC's will be sufficient.

--braindump

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

some thoughts

-Problem

• explosive growth in ai hardware sales and API usage is driving massive increases in compute demand, which directly correlates with higher overall electricity consumption and upward pressure on energy prices in the future.

• ev adoption adds significant new load to the grid particularly during evening peaks from home charging which strains capacity further and accelerates the rollout of time of use/tou pricing structures, making on peak electricity more expensive for households.

• cost of having local compute at your house is going to be astronomically high if you don't plan for this now you will forever be renting compute/api

-Opportunity

• hardware costs for compute continue to go up as more people realize they need ai to stay competitive

• battery energy storage systems for both residential and commercial will see explosive demand as people and businesses seek to arbitrage tou rates charging during cheap off peak hours or from solar and discharging during expensive peaks. this will cause supply bottlenecks which will push total system costs and market value much higher.

• solar panels and peripherals will be in heavy demand from both business and home owners pairing them with battery systems for self consumption and backup, leading to supply chain bottlenecks and higher prices

-solution

• you should plan on investing for local compute hardware and integrated energy/backup systems in place to have an edge where others will scramble to obtain said resources at higher prices and probably lower supply

--braindump

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

@WazzCrypto It’s a good marketing scheme to sell api

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

for reference on how much compute you need to replace an average employee/human

humans speak at around 3 tokens per second
humans write at 0.5 tokens per second
humans type at 2 tokens per second

small to mid weight models say around 7b to 13b parameters can run at 30 to 60 tokens per second on consumer grade hardware

you dont need a supercomputer to do the job of an average human that is not highly niche specialized when it comes to terminal jobs

--braindump

[Pandah (🥃,🦉)🐸🧙🏽‍♂️ @CRResearcher]

@thechadstack @rezoundous lol