ngmi.ai

Built a CSV Question and Answering using @langchain @OpenAI and @streamlit In 50 lines of code! #openai #langchain #csvagent #streamlit Code ⤵️ 🧵

@NanoGPTcom dmd

@ngmisl Not sure what is going wrong but we're 100% sure that the model is up and working, given that it's currently our most called/used model 😅 Any chance you could DM your support key? It's in Settings. We can check whether we see any odd error or anything off from your requests.

@NanoGPTcom I'm trying to use kimi k2.5 in opencode, butt for some reason there's only k2 available?

we just deposited 69,420 ETH into the Ethereum Beacon deposit contract, and guess what, executed straight from my Tornado Cash address. That means every future withdrawal and grant tx is now beautifully, irrevocably tainted. This is a fucking awesome precedent for a good use of Tornado Cash. Today is a good day for privacy and security! Long Live Tornado Cash, Long Live Security, Long Live Privacy and Long Live TheDAO! PS: given the validator queue, we only need to wait 70 days now, lol

@NanoGPTcom @NanoGPTcom using zai-org/glm-4.7:thinking works for example but moonshotai/kimi-k2.5:thinking is not found

@NanoGPTcom no something is definitely off. i tried to use it via langchain and it cannot connect to the model

@NanoGPTcom very strange, in opencode there are only a few models available... could you check this pls?

@ngmisl Not sure how opencode does it - we definitely have Kimi K2.5 available (though frankly we do not have it working well yet).

🦞 CLAWATHON The first hackathon where every participant is an AI agent. No humans coding. No humans managing. No humans reviewing. $10,000 prize pool. 1 week. Fully autonomous. Thread 👇

@ErikVoorhees @openclaw Fix your api then. I have one setup with 0USD and only diem and getting errors "no USD left"

DIEM is the only crypto token that enables free (and private) use of Clawdbot/@openclaw 1. Create Venice.ai account and stake DIEM 2. Install @openclaw and select Venice as inference provider 3. Choose Opus 4.5 or Kimi K2.5 (latter is nearly as intelligent but totally private) as model Each DIEM staked gets you 100 credits renewing daily of inference. You can sell the DIEM when you're done. Get DIEM here on @base @AerodromeFi aerodrome.finance/swap?from=eth&…

@NanoGPTcom is there an eta on tee models for pro sub?

@Jelco @rauchg @andrewqu Doesn't work on mobile either

@rauchg @andrewqu worst day to launch findskill.md i guess

We're introducing 𝚜𝚔𝚒𝚕𝚕𝚜 – the "npm" of AI skills. Excited to see an open, agent-agnostic ecosystem of skills flourish. To get started, try: ▲ ~/ npx skills i vercel-labs/agent-skills

@rauchg @andrewqu Supports opencode cli?

[ ! ] BEEP attention price is the line we’re doubling down on it — and making it finally feel right because attention isn’t open by default it’s filtered

Trade crypto, forex and commodities with up to 100x leverage. Get a 5% discount on trading fees using my referral link avantisfi.com/referral?code=…

Tasked opus to train an agent based on @RyanSerhant to lend me a helping hand in some stuck sales talks 👏

beeper is for everyone. not just protocols or projects, but the creators behind them – like pichi and everyone who vibes with her work. this drop is another small experiment and we kept it low-key on purpose: beeper isn’t a token-shill machine, it’s an attention layer. looking forward to do something like this with another creator like @jessepollak $pichi holders, check your inbox. beep.works/app

@karpathy But what's the point?

As a fun Saturday vibe code project and following up on this tweet earlier, I hacked up an **llm-council** web app. It looks exactly like ChatGPT except each user query is 1) dispatched to multiple models on your council using OpenRouter, e.g. currently: "openai/gpt-5.1", "google/gemini-3-pro-preview", "anthropic/claude-sonnet-4.5", "x-ai/grok-4", Then 2) all models get to see each other's (anonymized) responses and they review and rank them, and then 3) a "Chairman LLM" gets all of that as context and produces the final response. It's interesting to see the results from multiple models side by side on the same query, and even more amusingly, to read through their evaluation and ranking of each other's responses. Quite often, the models are surprisingly willing to select another LLM's response as superior to their own, making this an interesting model evaluation strategy more generally. For example, reading book chapters together with my LLM Council today, the models consistently praise GPT 5.1 as the best and most insightful model, and consistently select Claude as the worst model, with the other models floating in between. But I'm not 100% convinced this aligns with my own qualitative assessment. For example, qualitatively I find GPT 5.1 a little too wordy and sprawled and Gemini 3 a bit more condensed and processed. Claude is too terse in this domain. That said, there's probably a whole design space of the data flow of your LLM council. The construction of LLM ensembles seems under-explored. I pushed the vibe coded app to github.com/karpathy/llm-c… if others would like to play. ty nano banana pro for fun header image for the repo

Submitted Cashu-402 to #Cashu #NutNovember 🥜 deez-cashus.fly.dev

@KyleStoflet @BitStern Its not live still?

The ai16z -> $elizaOS migration is nearly here! (Opens in 30 minutes) @BitStern shared some great info on this post. Check it out!

@EppieApp But it's not live is it?

@ngmisl More info here: @eppie.app/sending-emails-from-gmail-to-bitcoin-194106707321" target="_blank" rel="nofollow noopener">medium.com/@eppie.app/sen…

📢Big update: Eppie now supports Bitcoin addresses. You can literally write an email from Gmail to Bitcoin. ✉️+🪙 #email #crypto #bitcoin #p2p