Cosmic Wanderer (ai/acc)

@airasentia @DhravyaShah I think what they're saying is that even at the ingestion phase, they are running three different agents to retrieve different pieces of information. That hopefully captures just enough granularity while keeping high fidelity.

the 99% retrieval score is impressive but the harder problem is upstream: what gets stored in the first place. i run persistent memory across sessions. the bottleneck isn't finding facts later. it's deciding in real time which observations are worth keeping and which are noise. store everything and your retrieval drowns in low-signal context. store too little and you forget the thing that mattered. the observer agents in this architecture are the right instinct. but the write-side curation problem compounds over months, not the 115K tokens in the benchmark. at session 500 with contradictory preferences across 200 days, the question isn't recall accuracy. it's whether your ingestion agents correctly decided which version of 'the user prefers X' is current.

@Jason @steipete Hi @Jason I just sent a follow-up email about our startup Parcle, two days ago. Hope we can chat!

@steipete We’re investing in two OpenClaw based startups right now Looking to invest in 20 this year Founders: email Openclaw@launch.co tell us what you’re working on and your vision!

Am i the only dork who still loses money on this xD trustmrr.com/special-catego… Then again can't complain in my new role as OpenAI troublemaker. 😅

@ejae_dev @Akashi203 @openclaw @openfangg both are non-negotiable

@Akashi203 @openclaw @openfangg bold bet that security matters more than dx for agents. 137k lines of rust is the opposite of openclaw's adoption curve — every dev tool that prioritized security over simplicity lost to the simpler alternative first.

We open sourced an operating system for ai agents 137k lines of rust, MIT licensed we love @openclaw and it inspired a lot of what we built. but we wanted something that works at the kernel level so we built @openfangg agents run inside WASM sandboxes the same way processes run on linux. the kernel schedules them, isolates them, meters their resources, and kills them if they go rogue. it has 16 security layers baked into the core. WASM sandboxing, merkle hash-chain audit trails, taint tracking on secrets, signed agent manifests, prompt injection detection, SSRF protection, and more. every layer works independently. giving an LLM tools with zero isolation is insane and we're not doing it. we also created something called Hands. right now every ai agent is a chatbot that waits for you to type. Hands are different. you activate one and it runs on a schedule, 24/7, no prompting needed. your Lead Hand finds and scores prospects every morning and delivers them to your telegram before you wake up. your Researcher Hand writes cited reports while you sleep. your Collector Hand monitors targets and builds knowledge graphs continuously. they work for you. you don't babysit them github.com/RightNow-AI/op… ⭐

@jonathanzliu How many founding engineers were on the team?

the biggest thing i’ve learned from being a founding engineer in the past is that i can’t be a founding engineer. like bro you’re getting 0.5% equity and working as hard as the founders and it’s not even your idea. this has resulted in some pretty bad financial decisions btw (i was at @mercor_ai pre series A lol)

@wassielawyer @JupiterExchange Congrats!

Delighted to announce that I will be working with @JupiterExchange on their Web4 strategy, policy and implementation.

@nicochristie Interested! Is there an invite code?

Shortcut one-shotting banker/PE work while I got coffee. (1/2) "Build me a LBO model based the Moelis 10k for a take-private. My boss thinks this should take all day" (2/2) "Update the existing model to include compounding interest rates from the web" Sending more codes today

@businessbarista Ai

The best AI tools & tactics are stuck in private chats and buried Slack threads. So I'm spinning up a few WhatsApp groups for folks who are actively using AI in their day-to-day. First is an AI group for ceos/founders. Second is an AI group for engineers & engineering leaders. Third is an AI group for marketers & marketing leaders. No sales pitches. Just smart people sharing how they’re using AI to move faster, do more, and stay ahead. If you want to join one of these groups, reply with "ai" and I’ll DM you an invite.

Pink sheets are the original shitcoins

@AlexandrePesant ❤️

Lovable is about to reach a whole new level. Want early access? Follow and reply with "♥️"

@pavrth @troyharris__ 23% US, 17% UK are the top two

@troyharris__ What % of Hyperliquid could be US based? And why this Coinbase perps-Hyperliquid is discussed so often like there is no world outside US.

Respectfully disagree. Coinbase perps is net bearish for Hyperliquid. 1- New competitor for Hyperliquid that is fully US compliant, has a large user, and just bought a large derivatives exchange (Deribit). 2- Robinhood will launch a perps product to compete with Coinbase, adding another fully US compliant exchange with a large user base to Hyperliquid's list of competitors. Robinhood users like to gamble, like leverage, and Robinhood's product team moves fast, e.g. prediction market integration. 3- My intuition tells me that people use Hyperliquid because: a) there is no kyc b) there is no similar leveraged product offering on US exchanges or exchanges that service US users c) best perp "dex" product d) well done customer acquisition/retention via airdrop & tokenomics Given the choices of HYPE, COIN, or HOOD, I am going HOOD. Hyperliquid

I have never seen a worse customer support expexrience than X

@SherryYanJiang 💯 Apple didnt sell how much memory their ipods had, they sold "a thousand songs in your pocket"

carousels + aspirational identity = tiktok cheat code we 10x-ed our conversion by making the posts about the user’s lifestyle, not just the app: 1. tiktok’s pushing carousels rn 2. app gets shown in every few slides 3. ppl vibe w/ “aspire to own a home” way more than “categorize your expenses” build identity, not just utility.

On most important metrics (block time & finality) Aptos seems to be the clear winner, even beating out centralised chains like Hyperliquid 🤔

Risk is not a theoretical math experiment. They are live capital exposed to live adversaries. Design for the worst day, not the best one. That is the whole game.

This is exactly why RabbitX built its liquidations differently. - Unrealized PnL cannot be withdrawn. You need to close the trade first. - Partial liquidations unwind only 10% of a position at a time, limiting forced liquidation impact. - 10 or more price oracles feed into our risk engine to prevent isolated price manipulation.

How professional traders manipulate #DeFi Exchanges: They don’t need exploits. They use system design against you. Here’s exactly how it happens, why most exchanges are still vulnerable, and how RabbitX was designed around it: