Sarfaraz Rydhan

@Mazariuss Got it. Appreciate it. Why is bid cap a better tool for this than cost cap?

@safoo Bid caps, if an adset was overspending unprofitably I would put our average 30day CPA as a bidcap on that specific adset. For us its working perfect. Its spending less, sometimes even nothing, but when its spends it is always profitable

There's a lot of noise right now about Meta spending on the wrong ads. I'm hearing it from everyone dealing with it. Here's the fix. I let Meta run my ad spend untouched for over a year. Right now it's the worst media buyer in my account. Since Q4 last year I barely touched anything. Let the algo decide where the money goes, full hands off. And for the last month or two, it's been getting it wrong almost every time. The usual response when I say this: "that's not Meta, your ads are just weak," or "the ad is TOF, it's just warming the audience, that's part of it." No. We actually tested it. We turned off the exact ads Meta was pushing spend into, and overall campaign performance went up. Not down. Up. So now we step back in, carefully. Two levers: → Bid caps on the bad adsets, so it only spends when it can actually hit our number. This is the main one. → Turning off the ads Meta over-spends on. We do this the least. Hands off worked for a year. Right now it doesn't. Watch what the algo spends on, not what it tells you.

@fourweekmba @icanvardar It’s really good and so accessible. Lives right where people are already asking their questions.

@icanvardar Not on the consumer side, billions of users are using it everyday; they call it Google AI Overview (and AI Mode)

google already lost the llm race no one i know uses gemini anymore, i don’t even hear it mentioned that often

@dams_app @camatkinsonnn Everything you mentioned is how I setup my meta ads this week: - Optimize for Trial Starts - 1 ad per ad set, 1 creative asset per ad - 4-5 ad sets in the campaign, equal budget on ad set level - US market only - kill losers after 3 days What do you think? 🤔

@camatkinsonnn Are you optimizing for ‘start trial/subscribe’ or ‘app install’ events ? How many campaigns? 1 ad set per campaign ? How many ads per ad set ? Only targeting US ?

A lot of people in my comments asked exactly how we scaled Pep AI Meta ads to push past $200k/mo. Most people burn cash on paid social because they just copy & paste basic tactics. Here is the exact framework I used. - The Creative Matrix: Standard influencer UGC actually bombed for us. The real winners? Ultra-short, 10-20 second hyper targeted UGC concepts and AI static ads designed specifically to disrupt the feed and convert high-intent traffic instantly. - The Testing Protocol: We don't guess, we let the data talk. I isolate our best concepts and feed them into dynamic creative testing (DCT) environments. The algorithm finds the exact winning combination of hook, visual, and copy before we aggressively scale the budget. - The Scaling Mindset: You have to spend money to buy data. Stop stressing over day-one profitability. Find the winning creative angle, optimize the funnel, and then go absolutely all-in. Drop your marketing questions below, or grab a 30-min Q&A strategy slot via the link in my bio!

@athcanft which path do you think works best: don't touch audience controls and enable all placements/enhancements and let Meta algo cook, or do you stay tight on audiences, your creative, and placements?

@jogicodes Do you limit how many accounts per phone?

This is how I try to target a US audience on TikTok step by step 🇺🇸 (Pro Tip at the end!) 1. Buy cheap used iPhones on facebook marketplace and do a full factory reset ($35–100) 2. Set up the iPhone: 🇺🇸 English (US) 🇺🇸 region United States 🪪 create new Apple ID ⛔️ turn off location ⛔️ tracking ⛔️ Siri ⛔️ Screen Time 3. Set up a residential proxy with fixed IP address 4. Download a proxy app from app store 5. Turn on the VPN ❓Check your IP address and location and set time zone city to be same as your location General → Date & Time → disable "Set Automatically" → set to YOUR (!!!) US city 6. Download TikTok and sign up with Apple ‼️ Ensure VPN is running beforehand ✅ Quick check: Do you get US content (President Trump, US sports such as baseball, football, basketball)? 7. Warm up account for 3-4 days by scrolling 10-15 minutes. Like relevant content in your niche. Don't overdo it. Less is more 8. Post content once a day from day 4 or 5. Check if it gets views. ✅ If it does, you're good. ⛔️ If not (views stuck at zero), start over. You made a mistake. Be more meticulous next time. Pro Tip: If you're abroad, make sure you scroll during US daytime hours. Set your lock screen to explictly display USA time.

‘How are you so focused under pressure?’ Me in 1995:

@Cuki98 @pcshipp Yeah. Needs a hard paywall. 500 users to download an app is no joke. Don’t give up just yet , see if they ones coming are willing to pay

@pcshipp Switch to a hard paywall or a free 3 day trial on yearly plan

It’s been 35 days, and I’ve finally lost hope in my first app - $0 MRR - $6 revenue - 520 new users (All users from ASO) It took me 35 days to fail my first app

@adriansolarzz Love this. Thanks for sharing the alpha.

@Pitt_13_9 @Sprouse_68 Now decides to throw down field with his punter

South Florida interim coach kicking FGs on 4th and 6 from the plus 30 in a Bowl Game … @Sprouse_68 “what are we doing?”

I love the robots I saw in San Francisco at @fdotinc today. Off to see an immersive movie in Apple Vision Pro. And just saw the latest Chinese Robots thanks to a video call, Am ready to talk robots with @rwang0 tomorrow.

@atbeme Amazing work !

Super excited to finally share some information about some of our customers Cal AI moved super fast, it’s awesome working with them and powering their inference They are proof that specialized models are finally viable at scale

@ishanmkh @rox_ai Amazing execution

Rox (@rox_ai) scaled to 25 million revenue agents for Global 2000 customers — all while still in beta. Today, Rox comes out of beta as the world’s first open, globally available revenue agent platform. Just as coding agents 10×ed engineering, revenue agents are doing the same for customer work. Humans are evolving from workers to orchestrators, while agents manage the end-to-end customer lifecycle. The winners — like @tryramp, @cognition, and @nvidia — have already transformed. Their frontlines are supercharged, their back offices automated, and their employees are being re-skilled to be AI-native. The laggards are falling behind – they need to transform now or risk irrelevance. Global 2000 boards want to transform now but incumbent CRM and SaaS tools lack the three essentials to deliver ROI fast: AI talent, full context, and enterprise readiness. In contrast, Rox delivers ROI within 90 days, charges only for agent actions, and takes on all forward-deployed engineering. - “Everyone needs an AI revenue operating system.” — Jensen Huang, CEO of NVIDIA - “I close more when I use Rox — that’s the bottom line.” —@max11free, VP Sales of Ramp - “We f*ing ripped out CRM, threw it out” —Tanay Tiwari, CIO of Puma Energy We exist to serve and support the AI transformation of global revenue — and starting today, we’re open for business. (article linked below)

@AIAdsApps Ads

Want Meta, TikTok, Google Ad Libraries of Top 90+ AI Apps? Repost and comment "ADS" and I'll DM this list to you (Must be following)

Tomorrow will be 2 July 2025, which is the mid point between 2000 and 2050. Everyday after that we are closer to 2050 and leaving 2000 further behind.

Totally agree, it is the coolest part. I think some folks read it as a dunk on the past, when it really just highlights how far we’ve come. Back then, getting even a single feature to work (at docusign scale) took a ton of grit and time. This new era of building has fewer constraints, but same creativity. That's the fun part.

weird how “you could build that in a weekend now” gets people mad as if that’s not the coolest part

😑

Full report for yesterday’s Google outage - not feature flagging a new code path - null pointer crashing a binary when reading empty fields from Spanner - remediation causing a thundering herd as there was no exponential backoff - which required manual throttling to recover from

to be clear coding with an agent is still pretty amazing. will only get better

@atbeme Can’t wait!

Reply below if you want me to DM you when this drops 👇🏾

I’m so excited for this weekend I’m going to write and share the first post about how I leverage AI to write almost 90% of my code I’ll be sharing two things: 1. My philosophy, principles, and strategies 2. A concrete starter repo that exemplifies an AI-optimized codebase