Victor Zhou (@victorczhou) profile picture

building agents @clay

San Francisco, CA
Joined April 2013
Photo by Victor Zhou (@victorczhou)Photo by Victor Zhou (@victorczhou)Photo by Victor Zhou (@victorczhou)Photo by Victor Zhou (@victorczhou)Photo by Victor Zhou (@victorczhou)Photo by Victor Zhou (@victorczhou)Photo by Victor Zhou (@victorczhou)Photo by Victor Zhou (@victorczhou)Photo by Victor Zhou (@victorczhou)Photo by Victor Zhou (@victorczhou)
Some of you thought that Inkling 2 Small was too big at 594B parameters. We heard you: Inkling 2 Mini is a looped transformer with just 26B parameters, small enough to run on your laptop, powerful enough for agentic work with a score of 20.1 on AA Index. thinkingmachines.ai/news/ink…
600
1,627
982
17,380
3,018,916
😆 𝐕𝐈
Today we are launching Clay's new global agent. To celebrate, we are giving GTA6 to every user who upgrades their @Clay plan before the weekend. Our global agent is a new long-running, outcomes-based agent. It analyzes, makes recommendations, and executes GTM work on your behalf. Now in beta. While Clay reasons across your live campaigns and TAM, makes you a plan, and runs your plays, you can rob banks and go for a long walk on the beach in Vice City. No need to call in sick. Let Clay do your GTM while you do GTA. Get your copy: clayhq.typeform.com/GTMGTA (Personally, I don't get the hype. But I was also the nerd in high school who sold his Nintendo Wii to pay for our team's debate coach.)
1
93
The @clay global agent is finally here! This is going to become The Way to build in Clay... we've got a lot more cooking
MEET THE NEWEST WAY TO BUILD IN CLAY. Our global agent uses your business context to plan a GTM play, then builds it using our Search, Audiences, Workflows, and Sequencer tools. > Turn a fresh signal into an end-to-end play in one conversation. @clay uses your account history to find strong fits, skip accounts you already own, and flag dead deals worth another shot. > Clay gets better every time, suggesting what to remember about your business. You choose what carries into the next play. > Your agent runs on the same data infrastructure trusted by @AnthropicAI, @OpenAI, and @Canva — with 200+ data providers behind it. Try it now in open beta!! ↓
7
157
Victor Zhou retweeted
The Max Agency podcast with @hwchase17 is back for a second season! Watch Episode 1 with @clay’s Head of AI @jeffbarg on your favorite platform: ⏯️ YouTube: youtube.com/watch?v=efBlbQah… 🎧 Spotify: open.spotify.com/episode/2qV… 🎧 Apple Podcasts: podcasts.apple.com/us/podcas…
1
6
4
34
11,522
An interesting finding from yesterday: 6.2 Sol is meaningfully better than 6.1 Sol at the trickiest GTM builds with @clay, and the strongest model of its class that we've tested on these tasks. It might be the best model to build on Clay with if you need a certain kind of intelligence...
1
4
1
18
1,422
We played around with 6.2 Astra too, and it doesn't seem much better than 6.2 Sol at using @clay. We're going to continue stress testing 6.2 Sol this week.
1
1
79
Also, anecdotally Opus 5.5 seems to be good for building with Clay, but we don't see this shine through in our evals yet. Still looking into that discrepancy!
1
49
Victor Zhou retweeted
Today, we’re excited to relaunch our official MCP connection with @OpenAI’s ChatGPT so Clay users can create end-to-end workflows without leaving their chat. This brings the ChatGPT MCP and the Clay CLI functionality to full parity. Clay users can now casually drop this into ChatGPT via the plugin (lnkd.in/gyUVvAMJ) ”Can you give me every CEO hiring for new marketing leaders whose company grew headcount 10% in the last 6 months? Find their emails and phone numbers and send to my CRM and sequencer to begin outreach.” Anyone on your team, your best builder or your new employee, can create new prospecting lists and enrich them with any data they need, without being a Clay expert. Check out the video overview from @danerwilliams on our team.
9
7
3
58
5,392
Victor Zhou retweeted
Coding agents make it easy to produce a huge PR before anyone has discussed the design. @sophiebits shares how we handle that at Clay, including when to send it back before reading the code and which decisions deserve a closer look. ↓
13
5
1
42
1,827
Victor Zhou retweeted
How does @clay keep a billion+ records fresh? Our cracked eng team built 2 pipelines on @ClickHouseDB (1 for quality, 1 for usage) that turned a days-long process into seconds. watch this <2m recap or dive into the tradeoffs & architecture decisions in the post 👇 Jump into Clay to search the dataset :) makes.clay.com/clay-original…
Our data engineers worked with @ClickHousedb to track quality across 1B+ company, people, and job records. Coverage history that took days to reconstruct now loads in seconds. ↓
1
1
14
608
Victor Zhou retweeted
A billion records. How do you know your data is good? Clay built its data quality system on ClickHouse, turning coverage investigations that took days into answers in seconds. Read the engineering deep dive → clickhou.se/4zdKViB
1
5
1
14
1,339
we use @ClickHouseDB for a lot of things but this dataset visibility use case is one of my favorites
Our data engineers worked with @ClickHousedb to track quality across 1B+ company, people, and job records. Coverage history that took days to reconstruct now loads in seconds. ↓
7
228
Victor Zhou retweeted
Introducing auto-mode for Claygent. It picks the AI model for each row of research, balancing answer quality AND cost. Now on by default for new Claygent columns: - 47% cheaper than sending every task to the most capable model. - <1 second to choose the model for each research task. - You can still pick your favorite model anytime. Super cool thing: We post-trained an open-weight model with @appliedcompute to handle the routing. It runs on their AC2 platform and handles hundreds of millions of queries a month. @jeffbarg, Vyshnavi, @rhythmrg, and @_brylee10 break it down below.
5
15
10
90
18,514
Victor Zhou retweeted
Introducing Claude Haiku 4.5: our latest small model. Five months ago, Claude Sonnet 4 was state-of-the-art. Today, Haiku 4.5 matches its coding performance at one-third the cost and more than twice the speed.
305
866
470
6,928
1,342,281
Victor Zhou retweeted
we’ve been using a grapevine-powered slackbot internally and i really can't imagine working without it. it’s like chatgpt, but with all your company knowledge and it's multi-player. - support asks how something works → it offers an easy-to-understand answer based on the actual code, not an outdated spec. - marketing asks when something is shipping → it pulls the launch date and the most recent project update straight from linear. - product asks about sentiment about a feature in beta → it summarizes real customer quotes from our customer channels and survey responses. - eng asks why we've made a particular technical choice → it combines code and our notion design docs to explain the what and the why. it can even auto-answer if it’s confident enough (no tagging required). and since this all happens in slack, everyone can learn or chime in, making the answers better over time. i truly didn't realize how much time i spent spend routing and answering questions until i stopped having to do it. grapevine takes 30 mins to set up and the first month is on us. give it a shot! 👇🏽
I’m excited today to share Grapevine, a system that makes it really easy for AI agents to search company knowledge across, Slack, Notion, codebase, and more. The first app we’re launching with it is a company GPT that works remarkably well, better than any alternative today.
2
2
12
1,077
Update: I joined @gather_town to lead an AI experiments team, and today we're releasing our first tool: Grapevine Grapevine is the first AI Brain that actually understands how you work We've been testing across early beta customers, and its been magical to see the early impact
I’m excited today to share Grapevine, a system that makes it really easy for AI agents to search company knowledge across, Slack, Notion, codebase, and more. The first app we’re launching with it is a company GPT that works remarkably well, better than any alternative today.
4
2
1
32
5,416
I've been on the team building Grapevine since the beginning, and in the span of a few months our internal company GPT grew from a hacky experimental demo into my most-used work AI tool.
I’m excited today to share Grapevine, a system that makes it really easy for AI agents to search company knowledge across, Slack, Notion, codebase, and more. The first app we’re launching with it is a company GPT that works remarkably well, better than any alternative today.
1
3
9
1,005
There's still a lot for us to improve in Grapevine, but to this day it remains the most useful company-context-aware Q&A system I've tried, and I've tried a lot of them (and am doing my best to be as unbiased as possible).
1
1
139