Unsiloed正在把企业里杂乱的多模态文档(带表格、图表、复杂排版的 PDF、扫描件等)转换成结构化的 Markdown/JSON,供大模型和 AI 代理直接使用,同时附带引用、定位框和置信度评分。
可以清晰看到文档层正在变成新的数据库,并在未来直接有 AGENT 撰写,审核,传递。
1.近期
AI 代理可以像查询数据库一样可靠地查询文档——高准确率的结构化提取 + 置信度评分,让它们直接拉取正确的数字、条款和表格,不再需要人类全程盯着。
2.中期
混合智能代理系统加入验证、交叉核对和有限自主行动。人类只处理例外情况。发票、KYC、标准合同、理赔等流程开始大部分自动运行。
3.长期
文档越来越多地“原生”结构化生成。AGENT把文档当作一等公民的数据存储(可读写)。人类审核大幅减少,只保留在全新案例和监管签字环节。
We’ve raised a $7.5M seed round led by Felicis so AI can read documents as easily as it reads databases.
Finance, law and healthcare run on documents, and not databases like any other industry.
Soon AI agents will read and write more documents than humans, and they’ll need to query them the way software queries tables.
We’re building the computable layer that makes this possible.
To do this at scale, we've brought together ML researchers and engineers from Google DeepMind, Cognition, and quant trading firms.
We use computer vision and dual-stream models to turn complex multimodal documents into formats LLMs and AI agents can understand and reason over.
If your team wants to turn messy documents into reliable, AI-ready data, book a demo:
unsiloed.ai/book-demo
1
32
Cloudflare Q3 收入 +39%,四年来最快。
但比数字更值得研究的,是 CEO 反复说的一句话:
"如果 agent 是网络的新用户,Cloudflare 就是它们运行的平台,也是它们穿过的网络。"
一家基础设施公司,怎么让这句话从工程师圈子,一路讲到出版商、企业高管、监管和华尔街?
2
1
1,217
跨境卖家不缺货,缺的是"货卖给谁、怎么说、找谁推"。
用Claude + AIsa 搭了 9 条 Agent 工作流:从 TikTok Shop 差评里找改款方向,从创意中心拆爆款广告,用爆款倍数筛 KOC,再让 AgentMail 一人一封去谈合作。
每一步都附了接口链接,照着就能搭 🧵
Robinhood 最新 Q3 财报反应月活一个季度少了 110 万,ARPU 却创了新高。用户在变少,钱在变多。这是一次主动的营销转型:从大众流量平台,转向高价值交易者 + 财富管理平台👇
2
4
1,822
Robinhood Q3 Crypo 交易数据下滑与 Bitcoin 价格交叉对比和预测分析 parisyxc.github.io/robinhood…
1
25
1688 看了很多,还是不知道先做哪一个?
我用 Karpathy 的 LLM Wiki 思路,搭了一个 Shopify 选品学习大脑:输入喜欢的品牌,研究产品、品类和增长,把灵感留下来,继续 brainstorm 自己的想法。可在 CHATGPT和 CLAUDE使用
插件已开源,文中分享三个实际案例和使用方法。👇
6
6
727
Brand Research Brain 已开源 🧠
输入一个喜欢的品牌网站,研究产品、品类和增长,把灵感积累成自己的 Shopify 学习大脑。
安装说明与使用方法:
github.com/parisyxc/brand-re…
2
111
Check out Muse, your personal AI agent. Redeem my code in Settings within 48 hours of joining and we'll both get 1 billion Muse tokens.
Code: XR9ICZ
muse.ai/join
1
1
124
MAC MINI = AI Infrastructure Seed Asset
Today we’re open-sourcing Lily, the local inference engine we built for hybrid compute in Perplexity Computer.
Lily is specialized for Qwen3.6-35B-A3B on Apple silicon, built so on-device compute doesn’t bottleneck Computer tasks.
Read more: perplexity.ai/hub/blog/optim…
1
114
If you jump in and out in different LLM and forget about where your conversation happens in which LLM, This is a super useful tool to locate and search.
Have you ever remembered the project, but forgotten whether the conversation was in Codex, Claude, or Kimi?
That kept happening to me, so I built my first open-source project with coding agents: AI Project Finder.
It creates a private, local search index for work spread across different AI tools. Search with whatever clue you remember: a project or client name, a prompt fragment, a workspace, or a filename.
Key features:
• Indexes local Codex, Claude Code, and Kimi Code histories
• Organizes results by session or project
• Returns you to the matching AI session and workspace
• Keeps the index on your computer
• Supports macOS and Windows
Here is a 26-second demo. GitHub is in the first reply.
#BuildInPublic #AIAgents
147
AI models may become abundant, but reliable physical execution remains scarce. The bottleneck shifts from “can the model reason?” to “can the system safely perceive, move, grip, lift, navigate, recharge, avoid humans, work 24/7, and be maintained in messy environments?”
That makes the execution layer a mix of robotics, sensors, industrial automation, motion control, batteries, safety systems, and deployment services.
Physical AI cannot act without sensing. The investable category includes LiDAR, depth cameras, machine vision, radar, tactile sensors, and software that fuses these signals into a usable world model.
This is especially relevant for autonomous vehicles, drones, warehouse robots, humanoids, mining automation, agriculture robots, and smart infrastructure. $OUST Ouster is one of the more direct public-market names here. It acquired StereoLabs to combine LiDAR with machine vision, explicitly positioning itself for robotics and “Physical AI” use cases.
1
1
217
MENU
this OpenClaw bot finds restaurants with ugly menus, rebuilds them as live web menus, and mails the owner a postcard...on autopilot.
here's how agencies can land recurring contracts with this system:
- scrapes every restaurant in a city in real time
- filters by review count + rating + last menu update + photo quality
- pulls the real menu items from the official site, PDF, or Google reviews
- samples the brand palette from the restaurant's own visual identity
- renders a 9:16 brand-matched menu, hosted live at a QR-accessible URL
- writes a personalized postcard referencing a real reviewer and a real dish
- mails it to the registered office addressed to the owner by first name
every step from discovery to brand-matching to outreach is automated.
reply "MENU" + RT and i'll send you a free guide so you can build this too
1
123
AGENT
this Adaptive agent uses GPT's new image model to run your entire creative + marketing team, on autopilot.
here's how any business can replace their whole marketing op with it:
- generates fresh on-brand creatives for every platform, daily
- writes and schedules the content across every channel
- replies to every comment the second it drops
- DMs every lead and turns cold interest into booked clients
- boosts paid ads from your best-performing organic posts
- doubles down on the creatives that win, kills the ones that don't
- tracks performance + adjusts strategy automatically, 24/7
every step from brand → booked client is fully automated.
reply "AGENT" + RT and I'll send you the full guide so you can set it up too
99
lovely idea
this OpenClaw bot finds $500k–$1.2M homes without pools, renders a pool into their backyard, and mails the owner a postcard showing the before/after, on autopilot...
here's how pool builders can close $50k+ deals with this system:
- scans satellite imagery for mid-market homes with empty backyards
- filters by lot size, sun exposure & recent ownership change
- pulls the homeowner direct from public records (not shared leads)
- renders a luxury pool dropped into their actual yard
- calculates build cost + home value lift for their specific zip
- generates a cinematic video of their backyard with the new pool
- prints a personalised postcard with the before/after + QR code
- drops it in the mail + hits them with retargeting
every step from sourcing to outreach is automated.
reply "POOL" + RT and i'll send you the full breakdown so you can build this too (must be following so i can DM)
105
I just use the /plan-ceo-review to figure out a marketplace or saas product positioning problem for my vibe coding product
Replying to @garrytan
gstack is available now at github.com/garrytan/gstack
Open source, MIT license, let me know if it works for you. It's just one paste to install it on your local Claude Code, and it's a 2nd one to install it in your repo for your teammates.
161




















