Cristian Constantin Olarasu

🤖 Creating a fully autonomous AI developer to write and maintain code is still the holy grail of software development. There are many brilliant teams currently working on this, and each of them have their own unique approach. The team at Cosine have made a massive step towards making this a reality. “Cosine achieves state-of-the-art results on the SWE-bench benchmark” They have achieved a score of 43.8% on SWE Bench, which was verified by OpenAI. What's even more incredible is they have managed to do this with a lean team and a relatively limited budget. Though it’s important to note Cosine isn’t looking to replace human developers entirely, but rather augmented them with human-like assistants that they can collaborate with on any kind of coding task. Yang Li + Sam Stenner + Alistair Pullen Read more about their collaboration with OpenAI 👇 https://lnkd.in/gu7xXVPn

21

(Real) Product Market Fit is a journey Too often I find founders are pushed to declare PMF early and start scaling prematurely as a result 😞 But… If you go back to the initial post from Marc Andreessen on PMF this is what he wrote: --------------------- You can always feel product/market fit when it’s happening The customers are buying the product just as fast as you can make it - or usage is growing just as fast as you can add more servers Money from customers is piling up in your company checking account --------------------- So where does PMF fit on the different “fit” ladder? 1️⃣ Founder Market Fit - validated problem? 2️⃣ Problem Solution Fit - a few happy customers? 3️⃣ Go to Market Fit - a channel identified, tested but not scaled? 4️⃣ Scale Market Fit - or actually scaling it so that money from customers is piling up in your company checking account? See what I did there? 😂 We've glorified the idea of Product Market Fit And I find there's a mental shift founders make when they declare PMF → we should spend more → we should raise more But both are VERY dangerous if it comes prematurely Lots of people around you are going to push you to declare PMF too early → investors → boards → peers And it also doesn’t help that we all desperately want to claim PMF But if you start ramping up your spend too fast too early - you likely won’t realize it for 6 to 9 months and by then you've burnt through most of your runway I saw it play out too many times as a VC And I've experienced it myself as a founder Here’s the truth: → Every journey to PMF is different → Every journey to PMF takes a different amount of time And that’s okay... As long as you manage your runway along the way 💪 -- ♻️ Repost to help a founder in your network Click 'visit my website' at the top of this post to register for my Seed to Series A session DM if you want help getting ready for your Series A

54

21 Comments

AI memory is fundamental for conversational AI and AI agents. It is critical for building personalized experiences and improving reasoning. Check out Letta’s course to understand how to use AI memory to enhance the customization and quality of your AI app.

15

Don't Be Fooled By DAU/MAU - It Hides Your Real User Activity! When founders want to see how engaged their users are, they look at the (Daily Active Users) / (Monthly Active Users) ratio metric. While it only gives a single, blended average. It doesn't show where users fall across the activity spectrum. So instead of relying solely on this ratio, what should founders analyze? Analyze the Power User Curve. It's a 30-day histogram which graph the number of days per month users were active, provides a more detailed breakdown of user engagement to identify and understand power users. 1️⃣ When The Curve Smiles A curve that looks like a smile - with high bars on the left and right - it shows that users are highly engaged and returning to the app daily. As the product can create more power users, the curve will shift further towards the right side of the smile. Facebook has a very right-leaning smile curve, with more than 60% of its MAUs coming back daily. 2️⃣ A Left-weighted Curve When the curve is left-weighted, most users only have activity one day a month. By this activity metric, the product has very few power users. That's okay, not every product is built on daily returning users. An investing product like a professional network like LinkedIn is bound to have a more left-weighted Curve. Low engagement simply means the company must extract more value from each engagement and create revenue with a business model that isn't tied to daily usage. 3️⃣ Trends & Cohorts Seeing how the curve changes over time will show if a product getting more engaging and why. A positive shift towards the right means a cohort of users is finding value in the product. This can illustrate how a new release or feature is bringing users back more often. Likewise, products like Uber or Thumbtack would want to look at cohort trends by geographic location to see how their product is developing in different areas and building network effects. 4️⃣ Customizing Your Curve To get the most from your curve, customize it to your business. It doesn't need to be viewed on the 30-day scale. A SaaS/productivity product focused on the workweek and weekly active users might want to look at their power users on a 7-day histogram, and the activity doesn't need to be app opens or logins. The curve can look beyond visits by using a deeper action that is more likely to show users getting or creating value. A publishing platform might want to look at “posted content” activity to identify its small contingent of power users. This is how YouTube could look at its creators, or eBay at its sellers. There's no silver bullet for how to look at user engagement, However, the Power User Curve visualizes user activity distribution, helping identify power users and understand how to monetize the product around them, rather than relying solely on blended average metrics. . Source: Andrew Chen (Partner at a16z) #startups #metrics #users #activity #MAU #DAU #fund #funding

25

1 Comment

founders: this is a good summary of why VCs reject you, though keep in mind that stellaris isn't an early stage fund. that is, it's just fine if you haven't figured out everything in the early days. but by the time you are raising series A, you need to have clarity on many fronts- Just don't wait for everything to fall in place before you startup.

9

Vishal Sachdev highlights the strategic integration of open source and proprietary tech in architecting tech stacks, developer ecosystems and resulting business models. The world class companies effectively balance value commoditization in open source and value capture in proprietary tech. This is the strategic challenge for companies operating in #blockchain and #AI. I believe those who operate at the intersection of blockchain *and* AI stand to win this strategic battle...

3

1 Comment

Part 2: What does defensibility look like in the GenAI (hype) wave? In my last post I argued that no, SOTA LLMs won't subsume a large portion of the use cases founders are building today. Unless LLMs have access to a domain's training set (as they do in coding but not say in M&A), it seems safe to bet that they won't invalidate an application's use case. Now I want to talk about the other defensibility challenge: How "easy" it seems to be to build an AI application on top of an LLM. For more than a year, we've been hearing the dismissive classification: "it's just a ChatGPT wrapper." This is a wildly misleading framing. It's like saying that all SaaS is a wrapper around cloud hosting, or MySQL. Adding LLM-intelligence to a problem can be either super easy, or super complex. The bottom-line here is that it comes down to: 1) How close is the "answer" the user needs to what comes out of an LLM naturally. 2) How fault-tolerant an application is. 3) What is the format of the answer that the user requires. If an AI-use case requires lots and lots of data/training/expertise NOT available to the public and future SOTA LLMs - it's probably defensible. If it requires zero hallucinations - it's probably defensible (it takes a LOT of effort to manage hallucinations--and typically it is about focusing narrowly on a core area of expertise) If its answers have to be formatted into fancy deliverables, integrated into IVRs, allowed to send emails to customers - it also helps defensibility. Clearly, there are super crowded AI spaces where defensibility is more questionable. But almost all of these crowded spaces (SDRs, SDEs, generic CS chat, etc.) don't have much complexity. They are often: - Not doing that much to enhance the output of the models - Functioning only in a "human in the loop" use case that can survive errors - Not that integrated into workflows There are, however, a lot of ambitious founders that are building defensibility by tackling challenges in all of these ways. Saying they aren't defensible seems to me to be applying an illogical double standard. In the pre-GenAI SaaS era, many of the big winners built semi-complex applications that had tons of user value and ran up the score on revenue. It was an article of faith that they were defensible. Why would this change when the user value has clearly greatly increased with the arrival of LLMs? #ai #saas #vc

9

1 Comment

Data is the new driver of the power race, not even AI. Prove me wrong. AI models will become faster — it’s inevitable. Inference, the number of variable capabilities, the amount of tokens, and the volume of requests will all grow. AI models will be versatile and omnichannel — working smoothly on devices (hi, Apple). This trend is evident this month with releases from OpenAI, Google, Microsoft, and Apple. It all will lead to much higher AI adoption rates, for sure. And likely, "whose model should I use for my case?" won’t be a question — as they all will reach roughly the same level. So, the right question would be — “what data should I use to get the best possible results in my specific use case”? Because no matter how good your AI model becomes, retail data is still in retailers' hands. The same goes for biopharma, for example. The same goes for every vertical you can imagine. Where precise predictability based on genAI and specific datasets can be incredibly valuable. I envision big competition between those who handle data and those who lead the distribution. Here is where the regulatory aspect becomes especially crucial. As AI becomes more prevalent and influential across various industries, there will be a growing need for balance between innovation and regulations — to ensure fairness, transparency, and accountability without slowing progress. Who will win, and how brutal will this battle be? Oh, it will be interesting to see. What do you think on this?

81

14 Comments

Are Dutch and European founders not ambitious enough? 🤔 I just read Techleap's latest report: "AI Scaling Challenges for Dutch Founders." (Okay, I got AI to read it and share the highlights with me. 😜) This quote caught my (& Claude's) eye: "𝘞𝘦 𝘢𝘳𝘦 𝘥𝘦𝘢𝘥 𝘰𝘯 𝘢𝘳𝘳𝘪𝘷𝘢𝘭 𝘪𝘧 𝘰𝘯𝘭𝘺 4% 𝘰𝘧 𝘋𝘶𝘵𝘤𝘩 𝘈𝘐 𝘤𝘰𝘮𝘱𝘢𝘯𝘪𝘦𝘴 𝘢𝘳𝘦 𝘱𝘭𝘢𝘯𝘯𝘪𝘯𝘨 𝘵𝘰 𝘳𝘢𝘪𝘴𝘦 𝘮𝘰𝘳𝘦 𝘵𝘩𝘢𝘯 10𝘔. 𝘛𝘩𝘪𝘴 𝘱𝘪𝘵𝘪𝘧𝘶𝘭 𝘢𝘮𝘣𝘪𝘵𝘪𝘰𝘯..." Almost every day, I see posts and jokes about European founders thinking too small. As someone who works with Euro-founders everyday, I have a different view. I see: - Massive visions - Global mindsets from day one - Founders ready to take on the world It's not the founders who lack ambition. They're just working within the realms of what's possible in Europe. That's why many of them get funded from and/or move to the US. 💰 The real lack of ambition? That's from investors and governments in Europe. 🇪🇺 I'm still working on an article, "The Case for European Optimism." And I think it will show that European founders are just as ambitious as their US counterparts. The real question we should be asking is: - How can we create an environment that allows founders to dream bigger and reach higher? 🚀 And yes, I know this quotes mentions deep ecosystem issues... --- ↳ What's your take? Are European founders thinking too small, or is it the ecosystem that's holding them back? 👇

8

1 Comment

Unpacking AI Reseach Product Manager role so you don't have to. 🥅 The goal of AI Research Product Manager is to make sure that researchers work on something that will ultimately drive business outcomes. 🥅 Researchers want to research, a PM wants to create impact. 📜 AI Research Product Manager should be comfortable to read and understand scientific papers and should have a good idea on what might come next. 🗿 The challenge of the role is often to find the intersection of user and customer needs and the things where the research space is pushing towards. 🧪 It is also extremely challenging for a PM to build credibility and interact effectively with scientists. After all, you are not a scientist. 👩‍🔬 To be efficient in this role you should be comfortable building prototypes by yourself. These notes are from my recent podcast with Guillaume (Synthesia / Google Search). Subscribe to my substack to get more product management insights: https://lnkd.in/eF76ZQqD

27

1 Comment

How to End a VC Pitch Meeting The end of a pitch meeting is as critical as the presentation itself. While many founders spend the bulk of their time perfecting the pitch, the last 10 minutes of the meeting require as much attention. Abruptly ending a meeting leaves investors uncertain or disengaged. A structured conclusion, however, can lead to a clear path forward. Here are three objectives every founder should aim to achieve when wrapping up a VC meeting: Define a Clear Next Step Never leave a meeting without defining the next step. Investors have multiple demands on their time. Without a defined follow-up, your pitch will get lost in the shuffle. Ask directly when the VC’s internal meeting will take place. If they say their investment team meets on Wednesdays, say, “Great, I’ll follow up with you on Thursday.” This tactic demonstrates initiative and keeps the momentum going. If the VC hesitates to define a next step, it’s a red flag. It indicates they aren’t interested in moving forward. Don’t waste time chasing uninterested investors—focus your energy on those who show commitment. Guide Their Evaluation Process Most VCs won’t know how to evaluate your company. It’s your job to guide them. Highlight the metrics that matter in your industry and set the framework for how they should assess your business. For example, you might say, “Retention is a key indicator of our success. Have a look at industry benchmarks and see how our numbers compare.” Or, “To understand our customers' pain points, I recommend speaking with players in this particular segment of the ecosystem.” If you have a supportive paying customer or an industry expert contact, feel free to suggest a reference call. By giving them a roadmap for evaluation, you help the VC align their assessment with the most critical factors that reflect your company’s potential. Set Terms for Further Information Exchange If the VC requests documents you didn’t cover in the meeting—such as a financial plan—this is a chance to secure another engagement. For example, you can say, “I’ll provide the pitch deck we just covered, and let’s schedule a time to go through the detailed financial plan.” This keeps the dialogue active and ensures that further engagement is tied to your next move. It’s not just about sharing data—it’s about maintaining engagement and moving toward the next step in the process. Each piece of information you provide should act as a bridge to further discussions. Ending with Control The key to ending a VC pitch meeting is to maintain control of the process. Defining a clear next step, guiding the evaluation process, and negotiating further engagement are all critical components of a successful close. By managing these elements, you ensure that the conversation doesn't fizzle out but instead continues to progress. Reserve the final 10 minutes of your meeting to wrap up effectively—because how you end the meeting might just determine whether or not you secure the investment.

30

7 Comments

I reviewed 218 Pre-Seed to Series-A pitch decks this year. Over and over, I see the same 4 mistakes: 1. Too Long Most decks run 18+ pages. Aim for 10–12. Cut extra slides on product features—no one needs a deep dive here. And skip detailed financial projections at this stage. 2. No Story Don’t just label a slide “Market Size.” Use your slide titles to tell a narrative: “We’re in a fast-growing, fragmented market worth $X billion.” When read in sequence, your titles should form a cohesive story. 3. Weak Team Slide Early-stage investors bet on the founders. Show only the key players (co-founders or founder + 1–2 execs), each with their most relevant, notable achievement. This alone can make or break your funding prospects—especially pre-revenue. 4. Missing or Unclear “Ask” Always specify how much you’re raising, how you’ll use it, and what milestone it’ll get you to. Save valuation talk for the actual meeting. Remember: • Warm intros are the channel. • Your pitch deck is the story. If the story isn’t compelling, even the warmest intro won’t matter. At Capfund Group, we specialize in IPOs and private financing for public companies, but we’ve seen these same pitfalls trip up founders at every stage. Don’t let your deck undermine your vision. Let’s create a narrative that actually gets funded. #PitchDeck #StartupFundraising #IPO #PrivateFinancing #CapfundGroup #VentureCapital #EarlyStageInvesting

2

"If we’re measuring velocity in days or weeks, it might seem obvious to directly tie a startup’s velocity to the number of hours you work each day. But startup velocity is far more than just hours worked (a nuance that many adherents to startup “hustle” culture get wrong). Velocity is not only about how hard you work, it’s about how you work. At a fundamental level, startup velocity is a measure of a company’s sense of urgency." https://lnkd.in/gV9Z_ift

9

4 Comments

Fundraising strategies that work: 1. Finesse 2. Brute Force 3. Just be a 2nd-time exited founder 4. Just be an early employee at Ramp, Stripe, Open AI (any of the 3-5 hottest companies in tech will do) 5. Just have a trust fund Idk why more people don't do number 3-5. In all seriousness, if you're a 1st-time founder you won't be able to raise like a second time founder so just accept that and focus on what you can do.

78

12 Comments

84% of customers want to be treated like a person, not a number. Yet most AI measurement systems ignore the human element. 🤖 Smart companies track these human-centric metrics: • Conversation complexity scores • Sentiment analysis trends • Personal touch indicators • Relationship depth metrics Stripe's approach: • Equal weight to efficiency AND effectiveness • Balanced scorecard methodology • Regular human interaction audits • Customer satisfaction correlation The best AI implementations enhance rather than replace human connection. Looking to balance AI efficiency with human touch? Let's discuss your strategy. #startups #founders #growth 🚀 🚀 🚀 Are you interested in the current state of AI in sales & marketing? I'm performing some research here at Mythmaker to better understand a problem I bet many of us have: measuring the impact of AI solutions in sales and marketing. And I need your help! What You Get: - Free personalized benchmark report compared to your peers - Early access to full research findings & potential spotlight features - Potential spotlight features to share what's working (and what's not) Quick Facts: - 5-10 minute survey - Focused on practical experiences - All responses confidential Help me out by clicking through & sharing this quick survey — https://lnkd.in/enKv52xT

1

I turned down £10M in funding. Everyone called me crazy (perhaps I was) But I'd seen founders lose their companies to investors who: 👉 Changed the vision 👉 Forced quick exits 👉 Pushed growth over sustainability (Also, they wanted me to ditch my existing seed investors- I felt a responsibility and still do to the people who had faith in me and my vision) At The Brew , we chose bootstrapping. Yes, it meant: 👉 Slower growth 👉 Tighter budgets 👉 More stress But also: ✅ Full control ✅ Better decisions ✅ Sustainable growth Plot twist: We now help other founders avoid the same pressure by offering flexible office space without long-term commitments. 🚀 Sometimes, saying no to money means saying yes to your vision. What would you choose: Fast money or full control? #StartupAdvice #FundingAdvice #EntrepreneurshipJourney

24

4 Comments

Testing Business Ideas: The Ultimate Validation Experiments Library 📌 Here's your ultimate validation experiments library: 1. Validation Experiments ↳ Test assumptions before you build ↳ Save time, money, and heartache 2. Market Engagement ↳ Get real feedback from real people ↳ Understand your audience before launch 3. Minimal Products ↳ Build just enough to test your concept ↳ Iterate based on user feedback 4. Feasibility Prototypes ↳ Prove it can be done ↳ Technical validation is crucial 5. User Prototypes ↳ Put it in front of potential customers ↳ Watch how they interact with your idea 6. Experiments in Production ↳ Test in the real world ↳ Gather data from actual usage 7. Recommended Tools ↳ Leverage the right tech for testing ↳ Work smarter, not harder Remember: Ideas are cheap. Execution is everything. But smart execution starts with validation. 💡 Don't waste months building something nobody wants. Test, learn, adapt. Rinse and repeat. That's how you build a business that lasts. Credit: Paweł Huryn , Founder @ The Product Compass (https://buff.ly/4cPChef) Totally Follow Him 🌟 #BrainDumps | BrainDump #90

774

109 Comments

Fundraising Realities The Bar Is High, and Here's Why Even in a year when my fund was closing 20+ deals, I personally only invested in 1-2 startups annually. Smaller VC teams?  They might only do 3-5 deals per year or skip investing altogether for months. This is why fundraising is so challenging.  You’re not just competing with other startups.  You’re competing to be the best founder an investor has seen in the past 6+ months. There’s no room for winging it.  You have to consistently prove you’re the top choice  - Across multiple months.  - Across multiple pitches.  - Across multiple rounds. 🔑 So, how do you become the one investors say yes to?  It’s not luck it’s systems. Here’s how to build one:  1️⃣ Get Introduced by Trusted Connections  The best way to stand out is to be vouched for by someone the investor already respects. Relationships matter. 2️⃣ Craft a Narrative That Makes You Unforgettable  Investors hear dozens of pitches every week. What makes you different? Highlight your expertise, vision, and traction with clarity and confidence. 3️⃣ Run a Process That Shows You’re Exceptional  Investors aren’t just assessing your startup they’re assessing you. A disciplined, transparent, and well-organized fundraising process sets you apart. 🚀 Fundraising isn’t just about asking for capital’s about showing why you’re a founder worth betting on. Great outcomes are built on great systems that make yours count.

12

2 Comments

If I asked you, “When someone turns in a work assignment, how accurate is it? 80%, 90%, 95% or perhaps 100%?” We don’t think this way about coworkers’ spreadsheets. But we will probably think this way about AI & this will very likely change the way product managers on-board users. When was the last time you signed up for a SaaS & wondered : Would the data be accurate? Would the database corrupt my data? Would the report be correct? But today, with every AI software now tucking a disclaimer at the bottom of the page, we will be wondering. “Gemini may display inaccurate info, including about people, so double-check its responses” & “ChatGPT/Claude can make mistakes. Check important info” are two examples. In the early days of this epoch, mistakes will be common. Over time, less so, as accuracies improve. The more important the work, the greater peoples’ need to be confident the AI is correct. We will demand much better than human error rates. Self-driving cars provide an extreme example of this trust fall. Waymo & Cruise have published data arguing self-driving cars are 65-94% safer. Yet, 2/3 of Americans surveyed by the AAA fear them. We suffer from a cognitive bias : work performed by a human is likely more trustworthy because we understand the biases & the limitations. AIs are a Schrodinger’s cat stuffed in a black box. We don’t comprehend how the box works (yet), nor can we believe our eyes if the feline is dead or alive when we see it. New product on-boarding will need to mitigate this bias. One path may be starting with low-value tasks where the software-maker has tested exhaustively the potential inputs & outputs. Another tactic may be to provide a human-in-the-loop to check the AI’s work. Citations, references, & other forms of fact-checking will be a core part of the product experience. Independent testing might be another path. As with any new colleague, the first impressions & a series of small wins will determine the person’s trust. Severe errors in the future will erode confidence, that must be rebuilt - likely with the help of human support teams who will explain, develop tests for the future, & assure users. I recently asked a financial LLM to analyze NVIDIA’s annual report. A question about the company’s increase in dividend amount vaporized its credibility, raising the question : is it less work to do the analysis myself than to check the AI’s work? That will be the trust fall for AI. Will the software catch us if we trust it?

59

10 Comments