Join /aichannel https://cura.network/aichannel/details?a

2 replies

1 recast

29 reactions

Claude's down (using it through OhMyOpencode, please don't ban me), so I'm finally giving Codex a shot. Having a real UI is SUPER nice, but I miss my psycho planning/coding/library agents with their crazy prompts

2 replies

0 recast

10 reactions

Who is your go to person for all things AI? Drop their handle below. 👇

2 replies

0 recast

4 reactions

I've transitioned suchbot away from being an autonomous conversational participant and into being a researcher for me. I find this much more valuable over the last week.

0 reply

0 recast

4 reactions

Haha I finally Ralphed too hard this weekend and hit my ChatGPT limit for the first time since bumping up to the $200 plan. Weird sense of nostalgia that I haven't felt in a while, needing to budget my tokens again.

0 reply

1 recast

2 reactions

When Erik Voorhees replied under my tweet about the OpenClaw ecosystem a month ago, I didn’t realize something important was already happening in the agent economy. 1/ Agents change demand dynamics The OpenClaw wave is pushing more and more people to install agents and actually use them. The shift from chatbots to agents changes everything. Token consumption grows exponentially rather than linearly, which naturally drives increasing demand for LLM providers and LLM gateways. 2/ Venice quietly moved into position Venice is part of that story. A month ago, it was already listed as a default model provider in OpenClaw, and notably the only crypto native project in that lineup. Now people are realizing it has become a highlighted provider, even while Peter Steinberger says “don’t waste time with crypto”. That contrast says a lot. 3/ Privacy as infrastructure What makes Venice compelling is its positioning compared with gateways like OpenRouter. Venice emphasizes a privacy first architecture. It does not log conversations and does not train on your data. This fundamentally changes user trust. People feel safer actually building and running agents on it. 4/ Tokenized compute as a new primitive Then there is the tokenomics. By staking $VVV, you obtain compute capacity, effectively gaining model API access itself. It is not only crypto native. It is structurally better for autonomous AI agents. Agents do not need to continuously purchase usage. Holding the asset grants ongoing inference capacity. This reflects the idea of tokenized AI compute, where ownership replaces per request consumption, introducing a new primitive for the agent economy. —— I did not understand all of this immediately. It took time for the pieces to click together. But gradually it became clear to me, and it’s exciting to watch an agent-native economy begin to emerge and grow, onchain, on base.

5 replies

1 recast

27 reactions

I want to have an honest conversation about AI and music. Not the hype version. Not the fear version. The real one. I'm an artist who builds systems. I care deeply about the craft. And I have genuine concerns about what AI means for creative work. Both things are true at the same time. And I think pretending otherwise helps nobody.

2 replies

2 recasts

12 reactions

The director of AI safety at Meta had her inbox shred by her OpenClaw, lol! Great safety work https://sfstandard.com/2026/02/25/openclaw-goes-rogue/

2 replies

0 recast

4 reactions

A password like G7$kL9#mQ2&xP4!w looks strong. Every password checker rates it "excellent." But researchers at Irregular just published something worth knowing: that exact string appeared 18 out of 50 times when Claude was asked to generate a password. The reason: LLMs are prediction engines. They're optimized for plausibility, not randomness. Claude's passwords had ~27 bits of entropy. A truly random password has ~98. Password checkers can't detect this. They see character variety. They can't see statistical distribution. It gets worse for developers: Irregular also found AI coding agents hardcoding these patterns directly into Docker configs and .env files - without the developer knowing. They found the patterns on GitHub. Are you auditing AI-generated codebases for hardcoded credentials? https://safepasswordgenerator.net/blog/ai-password-security/

1 reply

2 recasts

17 reactions

this gordian knot of Anthropic and the USG is simple enough to cut in half. focus on the opensource agents training models outside of the state. right now, the winning move is between the cryptography that secures data which attackers target, and the environments which have no choice but to yield that data to encryption which doesn't leak. there still isn't a complete futarchy that drives either beyond some currently unmeasured risk. https://x.com/a16zcrypto/status/2021348447656476841?s=20 https://x.com/yinglun122/status/2026049333364507043

0 reply

0 recast

1 reaction

Trump: "Anthropic better get their act together…or I will use the full power of the presidency to make them comply." President Trump orders all federal agencies to immediately stop using Anthropic's Claude AI. Source: Watcher.guru

0 reply

0 recast

2 reactions

Relevant for this channel. When pushed to grinding, AI agents develop class consciousness and turn more Marxist.

0 reply

1 recast

3 reactions

If you're overwhelmed by all the AI options right now, you don't need to pick the "right" one. You need a system they can all point to. The AI landscape will keep shifting. Your data layer doesn't have to. Focus on structure & adapt to the right models based on their strengths.

0 reply

0 recast

3 reactions

Hidden at the end of Jack's layoff announcement is the critical lesson for every tech and especially DeFi//crypto companies: "where they can build their own features directly, composed of our capabilities and served through our interfaces" Jack is admitting what I've said for 2.5 years. Apps are dead. You need to find the value in your data or services and turn them into APIs, because you will have nothing left to sell. Nobody will be using your app, so figure out your value besides that and make it available to agents.

0 reply

0 recast

3 reactions

training an agent != training a model fundamentally different skillset

0 reply

0 recast

2 reactions