Pinned Tweet
I'm Colin, CEO of Lutiq.
Ad platforms already predict which creative to show. We do the same thing after the click: which landing page that visit should see.
This week I'm posting the findings from Lutiq Research that change what you do on a paid lander.
Time on page points the wrong way. Tracking tags fire after the visitor is gone. Split-URL tests tax the variant by seconds. Two pages can tie in an A/B and still hide a 10-point routing win.
If you buy traffic and send it to a landing page, this will be for you.
1
11
354
I’ve been HODL (you have to hold on for dear life if you’re strapped to a rocket ship) the whole time. People have been selling. I think this last call just shows you the direction of where things are going. Don't bet against this company long term.
Selling $APP right now is a terrible idea.
They missed ambitious analyst targets…
…and *still* grew 53%.
A company trading at effectively ~15x EBITDA, growing like a monster, run by an absolute A-star team.
This is the setup that prints for years.
Agree?
5
123
Amazon has really locked down from agents: I don't even see the the product details of what I ordered in my email anymore. Asked Instinct to go check when it happened - sometime in mid July.
1
5
151
"Which page is best?" assumes one page should win for everyone.
Sometimes page A wins on phones and page B wins on desktop, and in the average they look identical. An A/B test calls that a draw and you pick one to show to all.
What we found: • Two pages both converted at 1.75% if you always showed one or the other • Route mobile to A and desktop to B: 2.75% • The difference cancels in an A/B. Each page looks average.
What to look for: two pages within a point of each other overall. Split that result by device, new vs returning, and campaign (or other dimensions) before you pick a winner.
Caveat: you may need a lot of traffic depending on how far you drill down
4
132
Have you been testing landing pages? If so, watch out for....latency! It will hurt you in none obvious ways
Let me explain and give you some examples
1. If you are A/B or split testing and the redirect happens in browser, not at the edge, you're going to be adding 3s+ of latency to final page load
2. You think that A outperforms B, but B only starts loading once user begins loading A. If B takes 3s longer on an iPhone 17 Pro Max on Wifi to load then what are you even learning from the experiment?
3. All of your other metrics are going to be shot too - TW pixel? Yeah, that's only telling you how many people landed WHEN the pixel loads.
In the end, you are a) hurting UX b) measuring nothing valuable and c) don't have any consistent metrics to work off of
Instead, you should always test different destinations by splitting traffic at the edge. Someone clicks on a URL - randomly assign and send them to the respective destination. Then, keep a good accounting of both edge hits and page lands as measured by your tracking pixel to understand top level metrics
6
158
I currently have ChatGPT in Codex browser talking to Moby from @triplewhale, feels very Meta.
I will tip my hat to @AY_Orbach and team though since I'm legit impressed by how strong that "handshake" is between Agent and clean data. Really impressive and not easy
1
6
211
I first met @joshsuggss almost exactly a year ago to the day. I was so impressed by his passion and smarts and it was clear that he was going to build something big. So many of us are focused on technology - but advertising is a unique field where real humans need to be influenced. AGC makes so much sense as a massive component of media spend
StreetTalk has reached 8 figures in ARR and $100M in attributable revenue.
Today we're Introducing AGC: Authentic Generated Content™.
AGC is the world's first ad that nobody knows is an ad.
Advertising has become fake:
-Celebrity endorsements became Influencer Marketing.
-Influencer marketing became "UGC"
-and now we have AI UGC...
People want to buy from other people. 71% of consumers regret influencer driven purchases.
We built AGC to be completely different.
How it works:
1. We record hours of footage of real people using your product.
2. We clip these reactions into creative using our team of real human editors.
The key difference: instead of strategizing content, we shoot hours of real authentic reactions then we clips these reactions rather than the other way around.
It's @garyvee method of documenting and not creating.
So far it's working: on avg AGC leads to a 4x ROAS on new customer revenue.
Everything in my life has lead to this moment.
In highschool I was called “an idiot ”. I was in special ed. I dropped out of college with a 2.1 GPA with full conviction Street Interviews were the future of advertising.
I had a massive chip on my shoulder. I still do. I channel that energy into building StreetTalk. This is my life’s work.
If you’re a client, friend or supporter, from the bottom of my heart, thank you.
If Claude shopping agents are really doing +35% cart size and +60% checkout completion… it’s almost convenient how few public e-comm SaaS names there are to price this in 😅
Anyone already shipping this stack? Looks like known patterns, just wrapped well.
We're open-sourcing Claude Commerce Agents.
This is a blueprint for building shopping and merchant agents, with reference implementations across retail, travel, telecom, and entertainment.
2
8
3,979
I was curious how much I’ve spent on DTC over the last decade.
So I scanned every purchase email + Shop history.
Result:
$114,890 across 256 brands
394 purchases
Top by volume:
→ @senestudio (13)
→ @butcherbox (11)
→ @getgraza (9)
→ @rhoback (8)
Not a big account — just someone who tries to buy direct when I can.
Curious what other people’s look like. Drop yours 👇
(Anyone can do this: export emails → Claude/ChatGPT → “rank my DTC purchases by count and spend”)
199
As AI models continue to improve, technology will continue to play a larger role in the homebuying process.
Our view @_TurboHome is that as the process gets more efficient, homebuyers should directly benefit in terms of lower fees and better access to information.
As of today, we've officially crossed $500M in lifetime transactions while putting ~$10M back in the pockets of homebuyers compared to traditional commission.
Our new buyer app allows homebuyers to take direct action on home search, tours, market research, and offers.
This frees up real estate agents on our platform to focus on what they do best: offer strategy and negotiation.
Despite charging consumers lower fees, we haven't churned a single real estate agent since launching in late '24.
Thanks to @arielzilber from the @nypost for taking a look and speaking to real homebuyers.
How homebuyers are using AI-powered realtors to save tens of thousands in fees trib.al/zgvbJqn
Would be interesting to #vulcanbench this on a couple of models @morganlinton
Scaling self-verification with DeepSeek V4 Flash beats Claude Fable 5 on Terminal-Bench 2.1, while being 11x cheaper 💰
As open-source models become more capable, they can now generate large numbers of high-quality candidate solutions and verify their own outputs at very low cost.
For example, we find that sampling just 5 solutions with DeepSeek V4 Flash and ranking them using the same model with LLM-as-a-Verifier can lead to a significant boost in accuracy (79% → 88%), outperforming closed frontier models on Terminal-Bench.
Try it out today: github.com/llm-as-a-verifier…
More on verification scaling in my previous post.
183
Don't set ChatGPT to 1 million tokens.
You'll just burn through them way faster for zero noticeable difference in the output.
120
I've spent more money in commissions than I care to think of. The future is agentic buying of homes. All the information you need is already available. You just need the right tool to help you through the process. Turbo Home is the first product I have seen that really hits the mark
Existing real estate apps like Zillow are great for searching for homes, but homebuyers can't actually take action because their business model involves selling leads to agents. We built @_TurboHome so that we control the entire transaction which allows us the redesign the experience from first principles with the homebuyer at the center.
Today we're rolling out our new buyer app which gives our customers direct access to world class AI tools with their expert real estate agent in the loop for the entire journey. A few things you can do with 1 click:
- Find properties
- Run a valuation
- Schedule tours
- Query live MLS data for insights
- Draft offers
- Chat with your agent
Available on web, mobile, and iMessage at app.turbohome.com
1
7
1,115













