Isaac (@IsaacMarks68) header
Isaac (@IsaacMarks68) profile picture

cynical , highly opinionated recruiter who believes people pissing on your shoulder is not rain

Republic of Mozambique
Joined June 2016
Photo by Isaac (@IsaacMarks68)Photo by Isaac (@IsaacMarks68)Photo by Isaac (@IsaacMarks68)Photo by Isaac (@IsaacMarks68)Photo by Isaac (@IsaacMarks68)Photo by Isaac (@IsaacMarks68)Photo by Isaac (@IsaacMarks68)Photo by Isaac (@IsaacMarks68)Photo by Isaac (@IsaacMarks68)Photo by Isaac (@IsaacMarks68)
Bro neon is no joke
1
1
19
You can build anything with enough context Introducing our open-source Clay alternative powered by context.dev
Introducing enrichable.io The open-source Clay alternative. → Access 1B people & 100M+ companies → Find emails, numbers, and company data → 1000 credits/month free → much, much cheaper than Clay Fork it. Add integrations. Make it just right for you.
26
10
161
10,851
Introducing UltraSearch for agents. Get search results with data snippets that matter including: → 250M+ people with job titles → 75M+ companies with funding → 150K+ stocks with live prices & more Supercharge every search with Firecrawl Alexandria. alexandria.firecrawl.dev/go/…
13
21
32
341
26,180
Finding leads is easy. Finding their email is where most lists die. Any database will give you 10,000 HVAC owners. Run them through an email finder and watch what comes back. Small local businesses are where providers have the least data, so finders end up guessing from name patterns. So we stopped trying to find emails and started collecting the ones businesses already publish. The insight is simple. Small businesses hand over their real email all the time. To the government when they register. To their trade association when they join. To the chamber when they sign up for the directory. Those emails were typed by the owner. Not guessed. Not pieced together from a name pattern. They’re just stuck in places nobody wants to dig through. Here’s where we dug: 1)The SBA Small Business Search. Free public API, a contact name and email for each firm, searchable by industry. 62,800 firms → 25,519 named emails at real business domains. 84% passed verification. 2) State and city open data. Contractor licences, business registries, health department lists. 13,057 verified emails. 307 association and chamber directories, most of them PDFs. Gemini Flash-Lite reads each one and pulls out every name and email for about $0.00003 a listing. The part that makes it trustworthy: 3) Every email extracted by AI gets a second read. If it isn’t really in the source document, it’s thrown out. One cheaper model we tested made up listings on a blank page. This step is why that never reaches a campaign. We keep only emails with a real person’s name on the company’s own domain. No info@, no Gmail. Google’s Gemini Decision model checks the company fits the ICP. It never sees the person or the email. Then everything is deduped against Clay and run through real verification workflows. The result: - 45,008 published emails checked. - 35,486 verified. - 79%, for about $5 in AI spend plus verification credits. All from sources the business itself published.And it works for any ICP that shows up on a public list. Swap the industry codes and the directory searches, and the same pipeline runs for dental practices, roofers, med spas, landscapers. The real question isn’t “who should we email?” It’s “where did this business already tell someone their email?”
16
1
33
1,276
I thought we were dead in the water. We’d emailed every company in the ICP. Construction, trades and medical practices, $1M-$75M. The list was done. So we built a new source and plugged it straight back into @clay. Here’s the setup: 1) Pulled the SBA small business registry and 19 open state registries (contractor licenses, business filings, health permits) 2)Had cheap Gemini models read 300+ trade association member directories, PDFs included 3)Ran every company through Google’s new Gemini Decision model to answer one question: is this actually our ICP? The decision model is what made this work. It takes a question and the company facts, then returns a yes/no with a confidence score. We tested it on 100 companies we’d already labelled by hand. It got 91 right. It costs $0.000075 per company. That’s 7.5 cents per thousand. We run it as a Clay workflow function: an HTTP call to OpenRouter, with the timeout under 100 seconds or Clay won’t take it. Every table can call it now, not just this one. Then Clay does what it’s best at: Dedupe against our Clay lead cache, so nobody we’ve already found comes back Look up revenue in Clay’s native dataset (search quota, not credits) Feed the qualified rows into the tiered campaign build Result: 35,486 verified net-new emails in the first round. The model spend was about $1.60 in total. Both brands are now loaded at 30K leads each. The lesson for me: when the database runs out, the leads haven’t. They’re sitting in public records nobody reads, because reading them used to cost a VA a month. Now it costs less than a coffee, and Clay is still the system of record.
9
10
3
107
9,228
Isaac retweeted
Jensen - " What's your name " She - "Fook me " 😂🤣🤣 I am dying.... Lmao
250
1,215
182
12,228
1,113,806
A churned customer said no software could build the list she wanted and she was right, but I solved it with $40 of Parallel Web Systems Decision credits and Clay. 4 steps on how we did it She churned so I had a call and asked what she was trying to do. She was shocked no tool would let her build a list of companies that sell Payment processing OR ERP OR accounting software OR accounts receivable solutions OR point of sale OR a bunch of other financial software solutions. No database in the world can tell her this because of how they are structured. So here’s how we fixed it. Step 1 Find every company that could possibly be a fit. Use Clay CLI to enrich their customers to find what industries, keywords, derived data points etc they show up with. Use that list to move to step 2 Step 2 Make an audience search with all of these filters. Intentionally this is going to bring in a mess of companies that absolutely don’t fit. Step 3 Use the Parallel decisions api. For .001 you can ask 100 questions in one request and get really great answers back. So we literally asked if the company in the input sold POS, payment processing, ACH, Accounts receivable, etc as multiple questions in the same request. Step 4 When they were selling that solution we’d have claygent find the page and a little more information on it so they could store that to their Audience. And now this customer has an always on database searching for new companies to qualify for ANY of their base industries or keywords and then they can have a workflow automatically check if the company is a good fit. They also can sync their hubspot to add new companies and filters to the audience as they close more deals. Zero maintenance other than making sure your Parallel wallet stays filled. Do people watch videos on here anymore or just scroll past? I can make one if you want to see the step by step.
15
1
44
2,538
Isaac retweeted
BREATHE WITH ME
53
148
26
1,376
265,205
Wow no one is safe
Introducing Claude Meetings: add Claude to your team’s video call and it builds docs, whiteboards, and other artifacts with you while you’re still talking. Starting to roll out in beta on Team and Enterprise plans.
26
Good idea thanks for sharing @ENowoslawski
For the culture, over 80k podcast guests, when they were on their last podcast and how many they are on that we can detect. Double up with something like @clay for finding contact info / custom claygents and @ExaAILabs's podcast data and go nuts. docs.google.com/spreadsheets…
1
59
Isaac retweeted
Full guide to GTM engineering > What the role actually looks like > The complete stack > How to get started You'll want to bookmark this
33
32
7
328
37,607
We're giving away a Hogwarts Legacy custom wrapped NVIDIA GeForce RTX 5090 to celebrate its official release with DLSS 5 🪄 To enter: 🟢 Share this post 🟢 Comment #RTXON Terms and Conditions: nvidia.com/en-us/geforce/con…
13
We're giving away a Hogwarts Legacy custom wrapped NVIDIA GeForce RTX 5090 to celebrate its official release with DLSS 5 🪄 To enter: 🟢 Share this post 🟢 Comment #RTXON Terms and Conditions: nvidia.com/en-us/geforce/con…
84,426
71,904
7,699
66,869
6,063,306
In November 2025, I had a chat with Adam Robinson that shaped the direction of our whole company. Me: I'm thinking about selling @heyreach_io. We are growing like crazy and we have 70%+ EBITDA. I feel like we'll get a really good multiple and now's the right time. Adam: These things don't sell. I spent so much time trying to sell my companies. But you probably won't listen to me, so go on, try it, and at least you'll get a lesson or learn how M&A works. Me: Yep, I'll try it. Adam: Once you figure out that you're not gonna be satisfied with the offers, do this instead: 1. Keep running and growing HeyReach profitably 2. Remove yourself from ops as much as you can 3. Use HeyReach's cash to fund other businesses that will make money In May 2026, I aborted the M&A mission. Adam was right. Last week, we launched @loopsterGTM, following the exact same playbook we discussed on that November call... Rest in peace my dear friend, mentor, and person I really looked up to! 🫶 You have done a lot in the business world, but you have done even more for a ton of entrepreneurs in this space. Especially for me! Your legacy will continue and you won't be forgotten!
14
4
167
7,701
One sourcing partner. Two ways to use the result. Take your RecruitCloud shortlist and contact candidates yourself, or add email outreach through our infrastructure. Your approval comes first. Tell me which role you want to fill.
19
Finding a candidate and finding a usable email are different jobs. RecruitCloud connects eight finders in a waterfall, with LeadMagic first and email validation after discovery. Candidate context and contact status stay together. We built it for you.
14
How it feels building a TAM of 1.16 million companies straight off public award data Free domain guessing found 251,397 of their websites before I paid for a single Maps lookup Guess the domains first, then let Maps clean up the rest at ~$25 a month for 300k lookups
7
5
69
5,060
Do not confuse a long candidate list with a good one. RecruitCloud checks must-haves, surfaces supporting evidence and flags what needs review. Jev, AI models and recruiter judgment work together. The output is a curated shortlist.
17