VP of Engineering, International Social Business Building AI agents that work beyond demos. Multi-agent systems · Memory · Evals · AI-native teams
Earth
Joined February 2026
- Tweets5.4K
- Following3K
- Followers1.2K
- Likes1.1K
Pinned Tweet
Your coding agent says “done.”
Who checks the checker?
I’m building Phanes: an open-source agent network designed around one rule: a completion claim needs independent evidence.
swrm dispatches work. evalid replays the proof. volto is the planned role & memory layer.
Same model. Same 6 selected hard SWE-bench Verified issues.
Issues resolved = all official tests passed:
swrm 5/6 · evalid 4/6 · plain agent 3/6 · Superpowers 2/6.
A small, project-run test: one run per workflow per issue, not a full leaderboard result.
@OpenAI @AnthropicAI @GoogleDeepMind — what should an agent prove before we trust “done”?
@Chi_Wang_ @joaomdmoura — I’d love your take from building multi-agent systems.
30 seconds. Claimed ≠ verified.
Projects + evidence in the replies ↓
#AIAgents #MultiAgent
美联航刚把 AI agent 挡在订票之外：怕 agent 拿着支付信息和里程账号乱买、乱改、乱花里程。西南航空反其道而行之，把航班库存直接开放给 ChatGPT。一边怕失控，一边怕掉队，agent 时代的第一道门，航司先吵起来了。
7
4
106
OpenAI 提议今天起停供 Cursor。提议不是定论，但信号很清楚：模型商不想再给应用层打工。Cursor 只有 5% 的流量跑 OpenAI 模型，断供伤不到筋骨，伤的是姿态——做平台的，永远怕上游收水管。
4
3
114
今天 Google 的 Gemini Decision 1 有点意思：专门给小决策打分的模型，一次能评 1,000 个问题，不生成文本只给概率。Agent 越来越像公司了，大模型干活之前，先有个小模型排班。排班这事，确实不该用最贵的脑子来做。
6
6
235
Gemini Decision-1 只要 150–250 毫秒做决定，256k 上下文。生成早就够快了，卡脖子的是决策——谁先把判断做到毫秒级，谁先把 agent 变成真员工。 xitter.twitterapi.xyz/osanseviero/status/212…
Gemini Decision-1 刚被扒出来：输入能吃 262,144 个 token，输出只给 1 个。专门干打分、多选、判对错的活。会聊天的模型已经遍地都是，会下判断的才刚开始稀缺。
1
3
106
OpenAI 把商家搬进 ChatGPT 了：认领主页、Business Hub、还能发 Business Agent。19 小时 443 万浏览，热度是真的。但 Agent 眼下只给美国电商，非美国商家先看着。入口换了，门槛没换。
1
1
67
ChatGPT is becoming a storefront, not just a chatbot. The interesting part is who the agent answers to: the business, or the person asking?
Now in beta: Businesses in ChatGPT.
Claim your business, manage your information, and get insights on how people engage with you.
Eligible businesses can also build a Business Agent to answer customers’ questions and help them explore what you offer.
openai.com/index/introducing…
2
1
107
等周四重置的人可以先领这个了。最怕的不是额度用完，是免费重置捏在手里舍不得用，拖到过期。
Anthropic just gave every paid Claude user one free full usage reset. the announcement has 2M+ views. most people will waste it.
per Anthropic's ClaudeDevs post (Nov 10):
- Pro, Max, Ultra and Team plans
- one full reset, across all models and limits
- use it anytime before Dec 10, 2:00 PM PT
- it's in Settings > Usage
- why: they fixed a few issues that used more limits than intended
how not to waste it:
- don't burn it the moment your short session limit runs out. that one refills by itself
- save it for when your WEEKLY limit is nearly gone. that's when a full reset is worth the most
- before you use it, turn Fable 5.2's effort down (medium is the default; low for simple edits) so the fresh limit lasts longer
- set a reminder for ~Dec 8 so it doesn't expire unused
it's one reset. spend it like one.
3
3
140