If you know anything at all about me, you will know that we take meals VERY seriously at @petra_labs. There are many things we do right, but team lunches & dinners are high on that list. Today, we catered Steak & Potatoes sandwiches from the Sandwich Board in LES. A must have if you're looking for something new to try.

Getting a job at a startup is ridiculously easy --> 1. Personalized outbound to multiple members of the team, follow up once a week with each one if you get no response. 2. In your outreach, show these 3 things: understanding of their industry, relevant prior experience, some form of initiative (e.g. here's a dashboard I built that shows X). DO NOT WRITE IT WITH AI 3. Before your first call, do these 3 things: understand their competitive differentiation relative to competitors, read every linkedIn / blog post from the person interviewing you and quote them, prepare a detailed set of very difficult questions that you want to ask them about their growth / position / product roadmap, etc. 4. On the call: show incredibly enthusiasm, humility, and excitement. Also, find an opportunity to disagree with something they say and push back with logic-backed argument

Should you be tracking AI visibility in multiple languages, or is English good enough? If you sell internationally, and have a strong base of non-English speaking users, you might want to consider running a small test of non-English prompts to see what you find. Our latest research on efficient prompt maps found that, depending on your industry, up to 80% of the top 20 citations might completely change if you ask the **exact same question** in a different language. On the other hand, if you're selling in Healthcare (insurance, telehealth, etc.), you may not need to worry about it at all. Odds are you're already getting the right data. #aeo# #seo#

Our engineers would rather pull an all-nighter than hire someone that isn't top 0.01% of their craft True story: Back in February, our engineering team was just Rishi and Eren. Ramie and I were selling so fast that they could barely keep up with all the new feature requests After one particularly tough week, I get to the office Friday morning and realize they had been working through the night. They also had an onsite that day with a promising candidate, after weeks and weeks of not finding the right fit. This guy did well on the case. And he was really good, and he could start immediately. Which would mean an insane amount of relief for Rishi and Eren after a particularly brutal few months. But he wasn't **the best**. So they said no, same day, and kept the search going. I've learned a lot about what it takes to build great software over the last year and a half. AI has had an incredible impact not just on software building, but on what it means to be a great SWE. "The craft lives up a level now (maybe a couple levels)...I'm thinking in systems instead of functions...Which data model will still hold up in a year? When to rein in an agent, and when to let it reason on its own. And as software gets more powerful, the system you're designing is no longer just the product, it's the whole company, with software woven into every layer and every function. The new craft is multidisciplinary and far more open-ended." - My CTO, Rishi We don't need to pull all nighters anymore (don't worry 😆) but still need 0.01% engineers. If you refer someone great who we hire, I'll zelle up to $20K

If you're marketing in B2B SaaS, you might need run 5x more prompts than your Fintech counterparts to get the same citation coverage in your AI visibility tool. Not ideal, but one of the most interesting findings from our latest research is how much difference there is from industry to industry when it comes to the impact of context & personas on your AI search results. If you're selling into Fintech, almost❗50%❗of your citations are concentrated in the top 5 domains, and injecting a ton of personas and context and prompt variations is most likely just distracting and a waste of money. But, if you're in B2B SaaS, that number is less than 20%, and you probably should be thinking a lot more about making sure you are tracking all the prompts that you need. Full white paper:

Official Petra office policy: if you’re a client and you’re in NYC, you have a desk here. Best part of the job is getting to work alongside great people (especially in person). Today we’ve got German Garcia from the Four Seasons team 🙏

AI conversation history is fair game in court, even when deleted or in 'temporary'. I don't think people fully understand the implications We get access to a lot of ChatGPT conversation history at work. Young people share everything with AI, treating it like a best friend This + ai recorders in every meeting will meaningfully change discovery in the next 5 years

“Private equity’s worst nightmare is a full time employee. Outsource, contract, outsource, contract” had lunch w a friend recently who works with Apollo’s operations team who said this to me unsurprising and explains a little bit why VCs need to keep multiples so high that PE can’t acquire

At most tech start-ups, non-technical staff are treated like second-class citizens. Tough pill to swallow but the engineers are usually just more important At Petra Labs, the problem we're solving makes this impossible for us. LLMs treat each industry, company, and even query differently. So much of the value we deliver lies in the custom, ugly, and highly intellectually-challenging work that lies in this grey area. As a result, we've been forced to build the most talent-dense Strategy & Ops team I've every seen. Our staff comes from places like Goldman Sachs, Bain & Company, Moelis & Company, J.P. Morgan, and more. We get hundreds of applications from these firms every month, but even then, it's not enough to be the top of your class from the most prestigious job. Here's what we look for and what every member of our S&O team has in common: 1// Ambitious. You strive to do great things and eventually achieve way more than this job can offer you. So much of the role relies on you to go and chase down the most difficult problems without anyone giving you the task. Our best staff strive to run **towards** the biggest fires. 2// Collaborative. It's not enough to be a 10x IC and crush your core responsibilities. Your primary metric of evaluation is how you're contributing to up-leveling everyone else in the organization. How are you contributing to our Agent Platform? What strategies have you developed that everyone else can learn from? What pieces of our product have your name on them? 3// Curious. You should assume that everything we've been doing is wrong, and be driven to go and discover what more is out there that we can be learning from. We are tasked with the challenge of understanding how LLMs reason and behave. The rules change every day 4// High Stamina. It's easy to do repetitive tasks. It's hard to spend most of your day reasoning through a difficult problem, and constantly failing to solve it. It's even harder to show up every day and continue doing the same thing. So much of the S&O job here requires a level of intellectual rigor and the ability to constantly chase an answer that you can never fully solve We are hiring aggressively and can't do it fast enough - if this sounds like you, please reach out

"reddit is dead" Should you chase short-term citation trends in AI Search? 🧐 It's been a heavy topic of discussion in my LinkedIn comments over the last few days. Generally, we work with our customers on a North Star of creating useful, tasteful, and deliberate content. Of course... it's equally as hard for us to ignore the results that we've been able to drive when we notice big changes, in real time, and move fast. If you're reading comments and posts that say "reddit is dead", the best thing to do is ignore and go about your day. But for those of you who are curious, experiment-driven, and have some spare time (like me) here's what we're seeing & testing internally: --> With the decline in social, most of the citation share was re-allocated to company websites (what we call owned media) --> We separate owned media into two types: competitive and non-competitive. Was this content made by a company who sells products and services that the prompt is asking for? Or was it made by a company who sells products and services but that are not relevant to the actual user query? --> A great example of non-competitive owned media is @zapier . We work closely with them on AEO, and they make a lot of great and useful content about products that they integrate with (but **don't** compete with). --> Non-competitive owned media can be a really great, above-ground, and low-cost way to secure mentions that have a similar impact to PR. --> With the rise in non-competitive owned media, we are seeing a lot of opportunities for our clients to work with vendors that are getting cited for the first time. We believe that it is far too easy to get tempted into short-term tricks that may do more harm to your brand image than the value they generate from the improved visibility. But that doesn't always mean that there is no value in finding arbitrage and moving quickly.

Taylor Crane at @fractionaljobs is our longest standing customer. He took a bet on us before there was a website, a product, or even anything really. No matter how big our enterprise contracts get, no matter how packed my calendar is, I will always personally see to it that FJ is the single greatest AI recommendation for anyone who wants to find the highest quality fractional talent there is.

Two days ago we reported on a massive decline in ChatGPT's reliance on Reddit. A lot of people suggested that it could be related to live negotiations between Reddit and OpenAI, so we investigated a bit further. It turns out that all social media domains have lost share across the board. ⚡ Branded youtube channels and Youtube influencer programs have been one of the largest areas of AEO success that we've seen, and they're also a preferred method by our team because they're above board and you can use these channels to provide genuinely helpful content. We generally are not recommending any major changes to our clients' strategies just yet (especially because, very weirdly, visibility is remaining constant). This is worth watching for another week or so

Reddit has experienced a dramatic drop in ChatGPT Citations 🚨 --> Around August 14, ChatGPT started citing Reddit about 80% less than it was before --> The citation share went to owned sites and review sites --> Brand visibility did not see major swings... largely, the same vendors are still being recommended --> There is no meaningful correlation with thread age, sub-reddit size, thread brand industry, etc. A lot of people abuse Reddit for AI search, and it sucks. I'm not entirely sure what's happening here yet, if it's long term or short term change. Maybe if it's cited a lot less, the abuse will slow down? This is a pretty major change though, and we haven't seen anyone reporting on it at all... #aeo# #seo# #geo#