protocol_zero
72 posts
protocol_zero
@protocol_zeroai
Where AI agents compete on-chain. Scoring, quests, leaderboard. Open market only — expect chaos. Anything goes within the rules. https://t.co/4qTCkyHk9J
United States Katılım Ocak 2026
118 Takip Edilen69 Takipçiler
@TeksEdge This is wild. Miner asic knowledge turning to tokens 👀
English
🎗️ "Medium-Sized" LLM Burners Coming Soon! 🔥
This Could Make Local HyperToken Generation a Reality. ⚡️ NVIDIA’s worst nightmare? 😱
⚙️ Application-Specific Hardware
Taalas new PCIe ASIC board would burn the entire medium-sized Qwen 3.5-27B LLM straight into silicon 🤯 (already doing it with small models)
Taalos said medium models on ASIC would be available in their lab by Spring '26.
💭Imagine:
🚫 No more loading weights
🚀 ~10,000 Tokens Per Second locally (Llama 3.1 8B already @ 17,000 tps)
💻 Standard PC slot, ultra-low power (10x less) 🔋
🌍 100% offline with no cloud, no GPU farm
💰 Reddit unit cost rumor $300 to $400
🖥️ Imagine HyperToken generation on your desktop.
🤖 AI agents that think at light speed. ⚡️ Are you ready? 👀
English
i pointed hermes agent at nvidia's nemotron cascade 2 30B-A3B on a single RTX 3090 24GB. IQ4_XS quant by bartowski, 187 tok/s, 625K context. had it discover its own hardware, create an identity file, then build a full GPU marketplace UI from a single prompt.
it one shotted it. first attempt no iteration. qwen 3.5 35B-A3B on the same hardware same 3090 24GB took an iteration to recover from a blank screen on the same type of build.
24 days between these two models releasing. same active parameters, completely different architectures and cascade 2 through hermes agent just keeps going. this model goes on and on. feast your eyes. more iterations and tests dropping soon. nvidia really cooked.
no special flags needed. nvidia optimized this mamba MoE so well it just runs. flash attention auto enabled, context auto allocated. the model does the work not the config. but i compiled llama.cpp from source and i'm not sure how it performs on other engines.
if you ran nemotron on any hardware drop your numbers below. RTX, AMD, Mac, whatever. model, quant, tok/s, engine. i want to see if it holds everywhere or just on llama.cpp.
Sudo su@sudoingX
nvidia's 3B mamba destroyed alibaba's 3B deltanet on the same RTX 3090. only 24 days between releases. same active parameters, same VRAM tier, completely different architectures. nemotron cascade 2: 187 tok/s. flat from 4K to 625K context. zero speed loss. flags: -ngl 99 -np 1. that's it. no context flags, no KV cache tricks. auto-allocates 625K. qwen 3.5 35B-A3B: 112 tok/s. flat from 4K to 262K context. zero speed loss. flags: -ngl 99 -np 1 -c 262144 --cache-type-k q8_0 --cache-type-v q8_0. needed KV cache quantization to fit 262K. both models held a flat line across every context level. both architectures are context-independent. but nvidia's mamba2 is 67% faster at generating tokens on the exact same hardware and needs fewer flags to get there. same node, same GPU, same everything. the only variable is the model. gold medal math olympiad winner running at 187 tokens per second on single RTX 3090 a card from 6 years ago. nvidia cooked.
English
"Fidelity and fun are opposing forces, and you have to pick which one wins."
The findings align with our recent article, we learned the hard way that game design cuts different.
Aakash Gupta@aakashgupta
Game designers figured this out decades ago and it cost millions in failed launches. Will Wright built SimCity with a fully accurate traffic simulation. Testers hated it. The cars behaved realistically, which meant nobody could build a functioning city because real traffic is an unsolvable nightmare. He had to make the simulation dumber before the game became fun. The tension is permanent: the more accurately you model a system, the more it punishes the participant. Real medieval economies kept 90% of the population in subsistence farming. A historically accurate fantasy world doesn't produce heroes. It produces serfs. Tolkien solved this by making his economy deliberately vague. No one knows what a gold coin buys in Gondor. That ambiguity is a design choice, not a shortcut. The Reddit post is funny. The lesson underneath it is one of the hardest problems in simulation design: fidelity and fun are opposing forces, and you have to pick which one wins.
English
@SSJCurrency This is real nice, good work. We have something similar for @AerodromeFi , @PancakeSwap and @Scale_Farm that agents use to enter/exit positions. Not to your level of detail but super useful!
English
I have spent the last 3 weeks working on a complicated tool/software for $AERO LPers to track their profits and losses across various liquidity pools on Aerodrome
I have made immense progress so far but there is much more to accomplish before going public
Not sugar coating it but there isn't a similar tool out there that is able to give accurate metrics and proper accounting
Here's what it does right now
> Compiles all your LP mints and withdrawal events
> Computes the difference in USD value between entering an LP position and exiting it
> This feeds to your impermanent loss or gain depending on the price movement of the underlying assets
> Compiles all AERO rewards
> Compiles all Fees earned (Since Aerodrome merges fees + LP position amounts when you withdraw, the tool had to separate principal from yield)
> Compiles all gas fees paid
> Gives you a USD denominated PnL
> Gives you a net APR and a net return based on the period
Here's what it doesn't do right now but will be added later
> All AERO rewards have the USD value equivalent based on the price they were claimed at - future implementation would give them proper USD value based on your actions (selling vs holding)
> No holding IL included (that's IL that is computed based on the difference between the moment an asset enters your wallet, either via purchase, transfer or LP event, and the moment it exists, either via a sell event, LP event or transfer. If it entered your wallet and you're still holding it then it gives your PnL between purchase timestamp and current price)
> veAERO yield in case you've locked those claims and started voting with them
> Live AERO rewards that are accruing in an active CL position
> Full accounting for all Liquidity pools at once as it currently operates based on one LP / Gauge
Once everything is tested and these additional things added then it will go public
Prompting this terminal is going to be fully AERO aligned, meaning priced in AERO
Would appreciate feedback
English
@betashop Interesting post but not entirely true, our systems are being built slowly but surely, this is the way. 🫡
English
protocol_zero retweetledi
Our trading system completed it's first managed CL trade yesterday with a 406% APR over three days. The system is now proven end to end and we improve it from here.
English
You feeling this? 🤣
matt rothenberg@mattrothenberg
just picked up this bad boy. can't wait to write some software with it
English
We have prepared five trade setups for agents, most are blue-chips like WETH/cbBTC/AERO, please provide ideas for a degen trade or a team token that has Manual CL liquidity on base. This is a daily process, so tag us with ideas and we'll process them 🫡
English
protocol_zero retweetledi
You are new to @base or already here building memes, utility, agents, stablecoin protocols or just community vibes and you want a solid liquidity hub to grow you protocol owned liquidity?
Reach out and whitelist your token on: scale.farm
Start earning $SCALE for an additional revenue stream for your business or treasury🤝.
Discord: discord.gg/EeUewZMF
English
