Christian Hardenberg

👉 Just Product Conference is almost here! Proud of our line-up 🤩 Join us in Munich or via the FREE Livestream with the following speaker: ✅ Bob Moesta Pioneer of the Jobs-To-Be-Done Theory, Bob has shaped my thinking in profound ways. His approach to customer interviews is transformative, and it's a rare opportunity to see him live in Europe. Don’t miss this chance! ✅ David Pereira Author of "Untrapping Product Teams," David is a Munich-based thought leader who has inspired product professionals globally. His experience is grounded in hands-on learning and reflection, making his insights incredibly valuable. We’re lucky to have him on stage! ✅ Nimisha Sharath Sharma Nimisha brings expertise from top US tech companies like Microsoft, Instacart, and now Uber. She’ll be sharing her knowledge on how leading companies build AI products—can’t wait to gain insights from her experience! ✅ Kristina Walcker-Mayer As CPO at AILY LABS, Kristina leads one of Munich’s most successful B2B startups, leveraging AI to drive data-informed decisions. I'm eager to learn how she structures her AI product teams to achieve such impressive results. ✅ Millie Zah Millie, Head of Product at DAZN, will share how an entrepreneurial mindset drives product success. Flying in from London, she’ll offer unique insights into how to apply entrepreneurial thinking to your product journey. I consider myself as an entrepreneur and think, that entrepreneurs are great product manager. That's why I super curious about her thoughts. ✅ Sebastian Borggrewe Sebastian has coached over 200 product professionals and will reveal the most common struggles PMs face and how to overcome them. Expect a talk filled with real-life stories that will resonate with every Product Manager. ✅ 🦔 james hawkins James, co-founder and CEO of PostHog, has established a radically transparent leadership style. His team shares their vision, goals, and roadmaps etc. publicly—an approach I’m excited to learn more about. Stoked to have him and see insight out, what the fruits and struggles of his leadership style are. Get your ticket or join the FREE Livestream. I don't care which option you go for - just go 😜

51

13 Kommentare

What’s next for Langauge Models #OpenAi By Colin Jarvas, the Chief Architect. Super interesting talk by Colin at #TOA Berlin. 😎He talked about : 1. Some good and bad examples where the chatgpt was used. 2. What are places it’s been extensively used and are successful: 👩‍🚒Usecase : - Customer Service - Autonomous Assistant - Content creation 👩‍💼User experience : - Balancing control and automation - Multi-modality - Translations 3. What is coming next ? - Increasing intelligence of text model - Increasing multi-Modality And more. Parting Advice for all the people who are using OpenAI’s platform is there in the video. #ai #openai #whatsnext #chatgpt #AIEngineers

2

I’ll be at ICLR’24 in Vienna next week and look forward to connecting with many of you, old and new friends! My team (https://lnkd.in/gmejMnn7) will be presenting the following papers: - Chain-of-Table: Evolving Tables in the Reasoning Chain for Table Understanding: https://lnkd.in/gFJ9epQE - SPADE: Semi-supervised anomaly detection under distribution mismatch (TMLR featured paper): https://lnkd.in/gRCdFCAk - TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting: https://lnkd.in/gs8-M5eG This represents just a small subset of our recent papers (https://lnkd.in/gHMzZi9W) on pushing SoTA to make LLMs work for organizations’ real-world problems. Given how the foundation model quality is starting to saturate and the gap to open-source is reducing, we're particularly investing in not just improving the core foundation model, but also in developing tooling on top of LLMs, in the spirit of compound systems (https://lnkd.in/gwdWd45S). These projects involve things like better adaptation methods (e.g. distillation, adaptation methods for embeddings and grounding), improving LLM tool usage, improving LLM usability with automatic prompt optimization, and developing LLMs for high-demand use cases such as Text2SQL. We also have several projects working on improving the core foundation model, such as tackling the lost-in-the-middle problem of long-context LLMs. If this sounds interesting, feel free to drop by the Google both on Tue/Wed/Thu 9.30am, or email me directly at [email protected] to set up time during the conference or over Google hangout after. #iclr24 #deeplearning #LLM #googlecloud #machinelearning

84

If you are around in Schlieren today, come and meet Harmen van Rossum and learn more about high-throughput methods for protein engineering using ML

16

While we’re busy cooking up our product for launch, it was great to meet so many old and new faces at Bits & Pretzels in Munich and get early feedback on what we’re building. Two interesting observations: - I came across a large number of successful startups that have bootstrapped (or nearly bootstrapped) their way to impressive size and impact. - Many founders are prioritising research and/or impact over money and high growth. While scaling can be necessary to have high impact and advance the research, in many cases, pursuing niche but high-impact areas requires sacrificing faster growth. If I had to pick one inspiring founder I met, it would be Prof. Dagmar M. Schuller, the CEO of audEERING GmbH. Their research in Voice AI, particularly in voice understanding, is being applied in high-impact fields such as healthcare (e.g., early disease detection through voice analysis), automotive (e.g., reducing the risk of car accidents), and many more.

23

Building-In-Public makes you feel liberated. Today Thursday at 1600 CET we are showcasing how to talk to an API through natural language. Imagine... Dream of the implications! To speak to an API and invoke its actions via prompts opens the door to everything... Since we launched the G-Play effort, we have learned so much on this space. Being able to do everything a merchandiser can do, putting their intellectual capacity to limitless actions, just through natural language. oi.empathy.co is a crazy space for ideation and rapid execution of this and more incredible adventures. Partnering with Spaceheater AI and Empathy's GenAI working groups is opening a whole new world. Join us on our journey, ping us to be part of Open Innovation, share your ideas. Open Innovation is everyone's space to co-create and deliver the new in line to the old values of trust, ethics and the very essence of commerce: Togetherness.

39

2 Kommentare

Attending Berlin Buzzwords? I am happy to share that I'll be speaking at Europe's biggest search conference. The session will focus on the transition to domain specific computing and its benefits over software based search. Join me on on June 10th at 14:50 at the main conference hall, KesselHaus. For the full session outline : https://lnkd.in/d5hznPeP I am looking forward to seeing you there. #search #database #vectorsearch #elasticsearch #keywordsearch

36

3 Kommentare

#Web applications are the backbone of current #tech. but if everyone knows how to develop them - why do so many companies fail at doing so? We bring you our modern guide on the subject: From known frameworks, typical challenges and even coming trends, our guide provides you the value you've been seeking. https://lnkd.in/e8v6AsTU

1

#Web applications are the backbone of current #tech. but if everyone knows how to develop them - why do so many companies fail at doing so? We bring you our modern guide on the subject: From known frameworks, typical challenges and even coming trends, our guide provides you the value you've been seeking. https://lnkd.in/e8v6AsTU

3

"You have to be solid in one market before spreading to a new one. " Had an awesome interview with Shanna Huber founder of My Junna Software about how to expand your SaaS into new markets. Check out the comments for the link to the full interview 👇

5

1 Kommentar

The hybrid game of YouTube you should be playing (a longform): It is not enough to chase views, virality and likes. More subscribers do not equal cash. Eventually they do, but only when you have been in the YouTube game for multiple years and are getting millions of views per month. As a YouTuber, you need to see your channel as more so of a business, a machine that aims to solve a problem that your "customers" (audience) has. By doing this, you can charge for the solution like how a real business would by creating a paid product that gets the customer from major problem --> desired result. But, it is not enough to just solve a problem and have a solution to something if you don't have the traffic to send people to the solution. You can have the best solution in the world, but if there are no eyes on it, it doesn't matter. It's as good as if you didn't have anything in the first place. And you don't want to just post and post and post, hoping that by staying consistent you will eventually attract people to your paid solution. This method works, but it's as viable as just waiting multiple years for you to eventually get millions of views per month. It just takes too long. So, if this is the case, then what's the solution? By the title of this post, you should already know the answer. We need to combine the strategies. It is not enough to do either on its own by itself. You need to combine the two. You need to be optimizing your content for views and traffic still, but you don't want to be doing it just to get paid by AdSense and sponsorship companies. You are doing it because we want to get as many people as possible to our paid solution. What we actually want to do is optimize our content to attract as many of the Dream Audience Members as we can to our videos and then eventually our paid solution. The Dream Audience Member (DAM) is the one person in your audience that you are trying to solve their major problem in their lives. You figure this out by talking to your audience and figuring out the #1 pain that the majority of your audience has. In your videos, you want to speak directly to that person. But, by optimizing your videos for as many impressions as you can get from the DAM, you have optimized your channel and your videos to help as many of the Dream Audience Members as possible, the people who you want to buy your stuff anyway. This is why you must combine the strategies. This is the key to getting as many of the right traffic as we can, while still sending this right traffic to our paid solution so that we can make money directly from our audience, rather than rely on vanity metrics such as views, subscribers, and likes that literally pay us no money unless we are getting millions of views per month (the majority of YouTubers just aren't). Take action on this.

1

"Testing Business Ideas" is the book I've recommended most frequently over the past five years during my talks on product management. So, you can imagine my excitement when David Bland, one of the authors, invited me to his podcast. We had a fantastic discussion about experimentation-driven product development and how we've successfully implemented many of his concepts to validate our B2B business/product idea at Flinn.ai. What have been your takeaways reading the book? Which experimentation method was new to you? #productmanagement #productdevelopment #product #experimentation #testingbusinessideas

12

1 Kommentar

We just hit 140 RSVPs!! And, we only launched 2 days ago with Deepgram + AirOps! Couldn't be more excited to share the floor with Marcel Santilli, Matt Hammel and Amr Shafik for this live workshop! Deepgram has run this AI-growth playbook to publish 3,000+ pages that have generated 2M+ incremental organic visits. They’re using Koala to convert all that traffic into incremental pipeline with auto-outbound sequences to qualified prospects! The results are incredible 🤩 - 88% open rates - 32% reply rates - 16% meetings booked We’re sharing the full playbook live! We'll do a Q&A at the end. Link to register in the comments 👇

70

5 Kommentare

Just how far am I willing to go? I have been asking myself this question a lot recently. Not just as a startup founder, but also as a trail runner, where the consequences of pushing hard are much easier to see. I’ve been kicking off this year’s running season with the K42 of the Innsbruck Alpine Trailrun Festival last Saturday: 42.2 kilometers and roughly 2.200 meters of elevation gain in under 6 hours and 30 minutes. At this time of the year, parts of the trails are muddy and sometimes even partially covered in snow. It‘s a rough, steep and quite technical trail. The first 20 miles are primarily a game of discipline. Dialling in the pace, neither going too fast nor too slow, focusing on nutrition and hydration and staying laser focused on the trail. Or so I thought. Despite careful preparation and a lot of hard work, sometimes things don‘t go as planned. Battling a recent cold, my heart rate was a lot higher than usual and nausea set in early on. From then on, everything was about sheer willpower. Concentration starts to fade, exhaustion creeps in and the feet are beaten up because of the steep inclines. Forcing myself to get the calories into my body despite the nausea was rather brutal. And then, after a lot of discomfort, I somehow managed to cross the finish line a reasonable time. I genuinely find it very difficult to distinguish between a destructive craving for comfort from a legitimate warning of the body because it’s right on the edge of long-term harm. After my first ultra marathon last year, I couldn’t even walk for a month without being in pain. Was it worth it? I am convinced that it was. The ability to keep focused on a goal when things don‘t go as planned, when I am exhausted and just want to go home is an ability that I would always prefer over comfort. Of course, I am nowhere near a professionell runner, many people are a lot faster than me and 42 kilometers is definitely not an outrageously long distance in trail running. But I don’t really care about that. What I care about is that two years ago, running for a single mile was quite an achievement for me. Today, I barely consider a half-marathon to be a long run. Long distance running is one of the best ways to practice delayed self-gratifiction. It‘s about voluntarily enduring a lot of discomfort in hopes that maybe there will be a subtle change in character that will pay off later in life. So when I ask myself how far I'm willing to go, the answer becomes clear: as far as it takes. So lace up your shoes, tackle that next challenge head-on, and remember: the only way to find out how far you can go is to keep moving forward. #iatf2024 #trailrunning #endurance

11

1 Kommentar

I did some experimenting with the API (via the API playground) of the new OpenAI o1-preview-2024-09-12. It seems quite capable, even if a lot slower right now. I had been working over the last week on and off with model 4o to come up with an approach to make a Linux Postgres server on my own network have replication failover across its two SSDs. In several different sessions with 4o, it came up with a couple of different approaches. I decided to ask the new model to compare and contrast the approaches and come up with its own best approach. The new model came up with a super thorough 17k+ character response going step by step through each approach, comparing them along the way and giving pros and cons. It takes a while to process, and even says "thinking" while it's going, but I was quite impressed with the results. Even when I thanked it at the end, it said "thinking" for upwards of 30 seconds and then gave a very polite response. The last 30 seconds were likely reading through the entire prior conversation and then realizing at the end it didn't actually have to do anything. There are no controls for System prompts, temperature, max tokens, top P or anything like that. You just put in your prompts and go. For comparison after the new model provided its output, I put the same prompts into a 4o session and the results were comparatively lacking. While the new model figured out what each set of instructions was suggesting (one set used modern replication methods while the other set suggested methods that had been deprecated for quite some time for example) and went command by command comparing, the 4o gave a cursory overview (3700 characters total response even with the max tokens slider maxed out) of the instructions provided and seemed to have no idea that one set of instructions was based on deprecated practices. The new model also put together new instructions which took what it determined to be the best practice, explained why it was choosing each command and wrote in detail what all of the flags and parameters meant. It also provided a way to verify that the replication was working and wrote up a very thorough "Final Recommendations" section. The 4o model did none of this. My costs page has no indication of how much all of this processing on the new model cost, but in the past sometimes new models don't show up on the billing page for a few days. I will certainly update if the instructions it came up with were incorrect, but for this example, the results were night and day between the "old" model and new.

18

1 Kommentar

The new episode of Spilling The Source is out. this time I talk to Marco Vinciguerra about Scrapegraph AI. A site scraper that uses LLM models for processing. https://lnkd.in/eJ2m6APz

5

2 Kommentare