Edison S.

Edison S.@EdisonSOLO_ · 1 天前观点输出

@stormapexxxx soon, ser.

AI：延续个人公开立场，强调长期主义、执行效率与行业方向判断。

Edison S.@EdisonSOLO_ · 2 天前观点输出

LFG 👀 https://t.co/gEGfg70z1W

AI：延续个人公开立场，强调长期主义、执行效率与行业方向判断。

Edison S.@EdisonSOLO_ · 3 天前赛道影响

I spent the first year of @projectsolo wondering if we were too early. The market for verified human identity didn't feel urgent yet. The AI agent economy was smaller. The bot problem felt theoretical to most people I talked to. Marc is right that too early feels exactly like too late from the inside. The difference is what you do while you wait for the market to catch up. But I believe the right time has arrived.

AI：偏主题投资表达，释放对基础设施与新叙事的偏多信号。

Edison S.@EdisonSOLO_ · 5 天前观点输出

We've made a huge mistake. In 2022, I sat in a review meeting for a data labeling project. Someone noticed that workers were finishing tasks in about half the expected time. Nobody saw a problem because faster completions meant lower costs. We shipped the data and moved on. Before I built Solo, I worked on systems that relied on human feedback to improve. Looking back, what surprises me is that no one stopped to ask the obvious question: How did we know the feedback was coming from humans? I don’t know exactly when that stopped feeling acceptable to me. I just know it’s the reason I started @projectsolo The SoloMission beta is now open. We need a new platform for verified human feedback. DM me if you wanna try it out.

AI：延续个人公开立场，强调长期主义、执行效率与行业方向判断。

Edison S.@EdisonSOLO_ · 8 天前赛道影响

The largest labeling platforms in the world know their networks contain bots. The reason it hasn't been fixed is structural. These platforms get paid per completed task. Verifying workers adds friction to onboarding. More friction means fewer workers. Fewer workers means slower fill times. Slower fill times mean clients choose another platform. The client gets the benefit of verification while the platform pays the cost. That’s why the problem persists. @projectsolo was built differently. Verification is part of the foundation because it has to come before the business model, not after.

AI：偏主题投资表达，释放对基础设施与新叙事的偏多信号。

Edison S.@EdisonSOLO_ · 17 天前政策影响

One design principle I keep coming back to: Identity systems should prove what they need to prove, then forget the rest. In 2015, the US Office of Personnel Management was breached. Among the data taken: 5.6M fingerprint records. Those fingerprints were collected for background checks — not for a permanent biometric database. That’s the failure mode of centralized identity. Data given in one context rarely stays in that context. It gets retained, copied, transferred, and eventually breached. You can reset a password. You can’t reset your fingers. This is why we’re building @projectsolo around a simple rule: Verify. Issue a proof. No “trust us to secure it forever.” The safest data is the data you never keep.

AI：偏政策推进，强调合规落地与规则明晰，对监管主线更敏感。

Edison S.@EdisonSOLO_ · 19 天前赛道影响

A researcher I spoke to a few months ago couldn’t figure out why his fine-tuned model kept preferring formal, robotic answers. After weeks of debugging, they found the problem. It was the RLHF data. RLHF is how frontier models learn what “good” looks like. Humans rate responses, and those ratings shape future behavior. But the labeling platforms are paid per task + Bots could complete more tasks than humans, across hundreds of accounts. The only verification required now was an email address. AI models are increasingly being aligned to feedback that may not come from real, unique humans. That might be the biggest quality control problem in AI today. We need a new platform to guarantee data quality. @projectsolo

AI：偏主题投资表达，释放对基础设施与新叙事的偏多信号。

Edison S.@EdisonSOLO_ · 23 天前政策影响

While preparing for our launch, we tried to break @projectsolo ourselves. So I tried to create a second Solo account myself using a different email and a different device (iOS for my 1st account and Android for the 2nd one) It didn’t work. The 2nd attempt was blocked at the biometric layer. Same face, same cryptographic identity, the proof is unique and tied to just ONE account. That’s what “one person, one account” looks like when it’s enforced by cryptography instead of platform rules. We already knew it would work. Watching it happen was different 👀

AI：偏政策推进，强调合规落地与规则明晰，对监管主线更敏感。

Edison S.@EdisonSOLO_ · 25 天前赛道影响

Exactly. A friend who runs a design agency told me he uses AI to generate initial concepts now. Faster, cheaper, more variety. But he's paying his best designer twice what he used to. When I asked why, he said: I need someone who can tell me which of these 200 options is actually good. People worry AI will make human work less valuable. What's actually happening is the opposite: When everyone can generate, the people who know what’s good become the most valuable people in the room.

AI：偏主题投资表达，释放对基础设施与新叙事的偏多信号。

Edison S.@EdisonSOLO_ · 26 天前赛道影响

AI can now generate a luxury brand campaign like this in minutes But brands are remembered because of how they make people feel, not just by how they look. That emotional connection is still one of the hardest things for AI to understand. I believe we don’t just need to build a better AI generation, we also need to bring real human emotion into AI. https://t.co/3iCkcoYFad

AI：偏主题投资表达，释放对基础设施与新叙事的偏多信号。