Andre Retterath
General Partner; Moderator
Sessions (8)
Sessions featuring Andre Retterath at Bits & Pretzels 2026 | Founders Festival in Munich
Welcome
The welcome session positioned the investor stage as a deliberately focused program designed to create maximum value for investors, institutional LPs, and, ultimately, founders attending Bits and Pretzels in Munich. The hosts opened by greeting the audience and explaining that this is the third edition of the investor program, hosted by Andre and Felix. They framed the third year as the point at which the format has become something of a tradition, while emphasizing that the original idea three years ago was not to introduce investors to Bits and Pretzels for the first time, but to create a dedicated program specifically for investors and LPs. Felix described the program’s development as an iterative process. In the first year, the goal was to bring top investors and LPs to Munich. In the second year, the team improved the format, and in the third year they again made “a lot of little adoptions” and changes. The stated purpose of these adjustments was precise: to provide maximum value to investors and LPs, while also supporting founders. Andre then expanded the current scale of the event, saying that more than 2,000 investors are present at Bits and Pretzels in Munich, with several hundred additional participants joining online because they cannot attend physically. He highlighted that this is the first time the audience has been extended in that way. A central point of pride was the presence of more than 330 institutional LPs. Andre described the effort to bring international LPs from all over the world to Munich as significant, and said the hosts were looking forward to the conversations ahead. He then set expectations for the two-day agenda: more than 60 speakers across more than 20 sessions, designed around deep and honest conversations. He contrasted this approach with conferences that rely on ten- or fifteen-minute panels, explaining that the investor stage is intended to go deeper into specific topics already known to matter to the audience. The welcome concluded by outlining additional networking opportunities and announcing a new report. Beyond the stage program, attendees were pointed to GP-LP matchmaking, the investor lounge, exclusive formats such as LP Fifty connected to Oktoberfest, and more open formats across the days. Andre then announced that the team had collected convictions from several speakers about the next five years and compiled them into a report published that day. Attendees can access the convictions report through a QR code in front of the investor lounge and later online. He closed by thanking the contributing speakers, the Bits and Pretzels team, and cohost Felix for making the program possible.
State of AI for VC
The core message of the session was that AI is changing venture capital not only as an investment category, but as an operating model for investment firms. The speaker framed the VC digitization journey as a spectrum: “old school VCs” using tools such as Zoom, MS Office, Teams, and Salesforce; “productivity VCs” using SaaS tools such as Calendly, Slack, Airtable, and Notion; and “data driven VCs,” split into workflow builders, full stack builders, and emerging quant VCs. Drawing on the Datadrim VC audience and a qualified landscape of data-driven firms, the speaker focused on how investors are using AI, automation, internal engineering, external data, and tooling to change sourcing, screening, diligence, and firm operations. A central distinction was efficiency versus effectiveness. Efficiency means achieving the same investment output with a leaner team, or increasing deal volume with the same team. Effectiveness means generating alpha and improving investment quality. In the dataset, about two thirds of VCs said they build for effectiveness, while one third primarily build for efficiency, with acknowledged overlap. The speaker then compared full stack builders with workflow builders: the median full stack VC firm had 23 employees, two engineers, nine investors, and about $800 million in AUM, while workflow firms had about seven employees, zero engineers, five investors, and about $150 million in AUM. Across 345 funds, the speaker highlighted a ratio of roughly one engineer for every five investors. The session then examined hiring, budgeting, and organizational ownership. Dedicated engineering headcount inside data-driven VC firms rose in the early 2020s, grew from 2023 to 2025, and began to flatten, while the total number of data-driven VC firms grew quickly. The speaker interpreted this as AI reducing entry barriers for funds to build internal stacks, especially among emerging managers and smaller funds. None of the most innovative funds planned to cut engineering hires; about half planned to keep engineering teams the same, and about half planned to hire one or more engineers. By contrast, only 2% planned to hire junior investors, and about 45% planned to cut junior investment roles, with AI cited as a causal reason. Budgets also shifted from roughly twice as much engineering HR spend as tooling spend to a one-to-one ratio between engineering HR and external tooling, especially tokens and data. The final portion focused on adoption patterns, tool stacks, impact areas, and risks. The speaker argued that senior technical ownership matters: more advanced funds tend to have a CTO, CPO, engineering lead, or data lead driving adoption, while less technically mature funds often rely on partners. Investment, sourcing, screening, and due diligence showed the highest adoption, followed by platform, talent, portfolio intelligence, operations, fund management, finance, fund ops, IR, and fundraising; legal and compliance had the lowest adoption despite heavy document processing. On tools, Anthropic had become the most used AI provider among Datadrim VCs, with nine out of ten using Claude; OpenAI followed, with three out of four using ChatGPT. Below the LLM layer, the stack remained fragmented, with Notion, Granola, Affinity, Etyo, and Harmonic mentioned. The highest reported impact was in deal sourcing and screening, while the top perceived risk shifted to data breaches.
“AI has reduced the entry barriers for investment firms to build their own tech stack.”
AI for VCs: Building the Data-Driven Firm
The core message of the provided transcript is that AI becomes strategically valuable for venture firms when it is connected to proprietary context, partner judgment, and workflows that help investors see, win, and support companies earlier and better. The panel opened with introductions from Moustafa “Mous,” CIO of Kleiner Perkins; Dominic from Headline; and Vic Singh, general partner at RRE Ventures and founder and CEO of Originalis AI. Each speaker framed AI for VC through the operating model of their firm: Kleiner Perkins as a lean but technically capable organization, Headline as a long-running full-stack data operation, and RRE as a highly collegiate partnership that relies on Originalis rather than a broad internal tool stack. The discussion then mapped the firms onto a spectrum between workflow assemblers and full-stack VC firms. Kleiner Perkins sits closer to full stack, with a dedicated engineering team covering data, data science, AI enablement, infrastructure, cybersecurity, sourcing, pipeline management, portfolio operations, and portfolio monitoring. Mous emphasized that despite Kleiner Perkins’ brand and AUM, the firm is lean: 57 people total, around eight investment professionals, and three engineers. Headline described an even deeper internal build posture, with ten to twelve engineers, around one petabyte of data, its own U.S. data center, Rust-based processing, and roughly fifteen years of internal memos, transcripts, emails, and other firm knowledge. Vic Singh argued that venture alpha should be understood through the sequence “seed, pick, win, build and harvest,” with judgment involved in picking and systems of work supporting building and harvesting. But he located the real alpha in “seeing and winning,” because once an opportunity becomes broadly visible as a deal, it loses distinctiveness. Dominic largely agreed and described how Headline uses internal and external data to track people movement, including LinkedIn data and benchmarks of individuals based on prior company trajectories. He argued that proprietary internal data is the only durable advantage when outside datasets are widely available. The panel closed, in the provided excerpt, by expanding on proprietary context and workflow leverage. Headline has digitized years of internal notes, calls, memos, and discussions into 1.5 million preprocessed “facts” that Claude can work with efficiently. Mous said Kleiner Perkins also treats context—board meetings, memos, investment history, call logs, and meeting notes—as the key data layer. He then shifted from “see it” and “win it” to “work it,” describing how AI can help a partner managing 38 meetings a week and 24 board seats by summarizing board materials into action items and pulling KPIs and OKRs into portfolio views.
Closing
The core message of this closing was that the day’s value lay in a deliberately human conclusion: after the earlier takeaways and insights, the speaker framed the shared lesson as “So human after all.” Rather than adding a new technical argument or introducing another framework, the closing distilled the preceding content into a simple audience-facing reminder. The speaker said they “love these takeaways,” signaled agreement with the conclusion, and emphasized that some insights had resonated strongly, including a reference to “The US tier one university.” The comment was brief, but it positioned the end of the session as an affirmation of ideas already surfaced during the day. The speaker then turned that reflection into a practical instruction for attendees. Using the German expression that “everyone’s just cooking with water,” they suggested that people should not treat others in the room as unreachable or fundamentally different. The concrete call to action was to “go out to the people and make some connections.” In the context of a closing segment, this was the session’s clearest value proposition: attendees should convert the day’s themes and atmosphere into actual conversations, relationships, and side-event interactions rather than leaving the insights abstract. After summarizing the spirit of the stage, the speaker gave logistical guidance for what came next. Attendees were directed to another stage, where Trevor Noah and Jokr Wintershteit were scheduled to speak together about “the price of the spotlight and the power of wit.” This announcement served as a transition from the current stage to the next programmed moment, while also keeping the tone energetic and socially oriented. The mention of drinks at various side events reinforced that the evening was intended to continue through informal networking and conversation. The closing ended with a precise instruction about the following day: the program would restart “tomorrow” at 09:30 sharp. The speaker closed warmly with “Have a good night. Ciao.” Overall, the segment functioned as a concise bridge between reflection, networking, the next headline conversation, evening side events, and the next morning’s schedule.
“go out to the people and make some connections.”
From Human Intuition to Superintelligence
The session’s core message is that the rise of increasingly capable AI forces a practical boundary-setting exercise: deciding when to trust machine prediction and when human intuition, values, and responsibility must remain decisive. The opening uses a vivid time-machine scenario set in 2040, where AI has surpassed human intelligence but daily life still presents irreducibly human choices. A parent is told there is a 40 percent chance of bullying on a child’s first school day; an AI can extend a dying dog’s life by six months, though the owner senses the dog does not want that; and an AI recommends a New York job opportunity with a “90977% probability” of personal and career success, while another option later appears more fulfilling. These examples frame the panel’s question: where is the boundary between AI, human intelligence, and human intuition? The panel is introduced as Noah Hollmann of Prior Labs, Matthias Niessner of SpAItial AI, Tim Rocktessner of Recursive, and Andre Retterath of Earlybird VC. The discussion then turns to recent claims attributed to Anthropic researchers that self-improving superintelligence could arrive this decade and that there is a nontrivial chance it could ultimately kill humanity. Tim Rocktessner responds skeptically, calling a claim such as a 10 percent chance of humanity being dead by the end of the decade a “massive overreaction.” He argues that extraordinary claims require extraordinary evidence, and that while AI-related cybersecurity and biosecurity risks are sensible concerns, the evidence cited so far does not point to a short- or mid-term future in which everyone dies. He also questions probability estimates from figures such as Jeff Hinton when they appear to rest on gut feeling rather than transparent evidence. Matthias Niessner similarly says the extinction framing is overstated from a technical perspective, while emphasizing that AI can still be dangerous depending on human use. He distinguishes between an intrinsic will in AI systems and humans deploying AI for harmful purposes, including defense applications where the key issue is who controls the trigger. Andre Retterath adds that even small percentages of humanity-level or severe downside risk deserve discussion because the technology is evolving rapidly and can create both good and bad outcomes. The panel agrees that safeguards, pressure on companies, and safer system setup can reduce the chances of things going wrong. In the available excerpt, the conversation closes by defining superintelligence: Prior Labs’ work on structured and tabular data is described as already superhuman in a narrow domain, while broader superintelligence is framed as a system able to solve all, or the vast majority, of meaningful human tasks over time.
Fireside Chat: Black Forest Labs x Earlybird
The core message of the provided session excerpt is that Black Forest Labs’ rapid rise came from combining frontier visual AI research with unusually effective distribution and a tightly knit technical founding team. The host frames Robin Rombach as one of the people behind Stable Diffusion, describing that work as helping launch the open image generation wave, and then positions Black Forest Labs as a Freiburg-headquartered AI lab with global impact. Earlybird’s role is introduced as an early backer, while the conversation focuses on how a small team moved from academic breakthroughs into commercially successful image generation models such as Flux, with ambitions mentioned by the host around video, world models, and physical actions for robots. Rombach explains that the Stable Diffusion breakthrough emerged from a convergence of algorithms, data, and compute. He traces the work back to his PhD period at LMU Munich with his now-cofounders, highlighting latent diffusion as a new way to train generative models on visual data including images and videos. He is careful to note that the early models were trained on relatively small compute compared with today’s systems, but the team could already see that increasing compute by factors such as eight or thirty produced meaningful improvements. This scaling evidence helped justify moving beyond the university setting into an environment with more resources, including the period at Stability AI. The discussion then turns to why Black Forest Labs reached commercial scale so quickly after its 2024 launch. The host cites a public figure that within one year after the first models launched, the company crossed $100 million ARR. Rombach attributes the early traction to the release strategy for Flux One: a mix of open-source release and proprietary API access to the strongest models. In his view, the open model created initial distribution for a team of roughly ten to fourteen people, while the developer ecosystem that formed around the open-source model helped funnel attention and usage into the company’s API products. Finally, Rombach describes the company’s formation as a product of long-standing working relationships across Heidelberg, Munich, NVIDIA, Stability AI, Freiburg, and the United States. The founding group had already learned how to collaborate and had strong technical depth, which he presents as an ingredient in early execution. On location, he rejects a simple Europe-versus-US framing: Black Forest Labs started with people in both Germany and the US, kept Freiburg as a natural center of gravity, and aimed to show that state-of-the-art models can be built wherever talent, compute, algorithms, and team quality come together. The supplied transcript ends at 7:19, so later parts of the 19:39 session are not available for content-specific analysis.
Fund I Was Easy. Now Raise Fund II!
The core message of the provided transcript segment is that raising a first venture fund and raising a successor fund are very different challenges, especially in a market where capital is concentrating sharply among a small number of managers. The moderator frames the discussion around a bifurcation in venture fundraising: in US data cited on stage, about 60% of capital invested into venture capital funds is raised by 10 funds, while many emerging managers sit in market for around sixteen months when fundraising, counting only those who successfully close. The panel is positioned as a practical look at managers who have raised Fund I and are either preparing for or experiencing the move toward Fund II or Fund III. The speakers represent three distinct angles on emerging manager fundraising. Michael from Allocator One explains that his firm raises fund-of-funds capital to anchor first-time managers globally, mostly at Fund I and Fund II. Aashna Kumar from Upside Ventures describes a London-based $30 million Fund I focused on consumer tech and AI, with a differentiated LP base that includes major creators with more than 500 million followers and subscribers collectively; she says the firm believes distribution is value creation and has introduced hundreds of thousands of paying retained customers to portfolio companies. Magnus from Left Lane describes a growth equity firm investing mainly at Series A and B, headquartered in New York, with more than $3 billion raised over six or seven years. Magnus explains that Left Lane’s Fund I was not a typical blank-slate emerging manager raise because the team had previously run a similar strategy successfully at Insight Partners for nine years, with full track record attribution and examples including Delivery Hero, N26, and HomeToGo. He emphasizes the value of warm introductions from trusted industry relationships and recommends building a broad, diversified LP base across family offices, institutions, and endowments rather than relying too heavily on one cyclical constituency. On spin-outs, he stresses avoiding direct competition with the former firm, communicating carefully, and preserving long-term relationships where both sides can benefit. Michael then describes Allocator One’s launch as driven by necessity after he and co-founder Felix ran out of personal capital while investing in managers from their own pockets. Their first fundraise began in fall 2023, a difficult moment for an emerging manager fund of funds, and ultimately drew mostly high-net-worth individuals writing $250,000 to $1 million checks plus a handful of large professional allocators. Aashna closes the visible segment by outlining Upside’s Fund I: talent LPs contributed about 25% of capital, while the rest came from institutions, family offices, and opportunistic high-net-worths. She says founder Jamie took more than 500 initial meetings, over 150 flights, and visited three continents, with roughly 90% of LP prospects previously unknown to him. The transcript cuts off as she begins explaining the importance of listening to LP needs and meeting them where they are.
Closing
The core message of this closing segment was that the second day of the conference’s investor stage had formally wrapped, while the event still offered value through continued in-person attendance, online access, recordings, and upcoming networking. The speaker thanked attendees directly and framed the close as both a conclusion to the investor-stage programming and a handoff to the remaining conference activities. The most specific value proposition was that attendees, including those who joined virtually, could continue accessing content through the recordings available in the Dataderm VC community, with the newsletter positioned as the route to that material. The speaker first emphasized the scale of the investor audience at the conference. They stated that more than 2,000 investors were present and added that “everyone is still around,” signaling that the event community remained active even as the investor-stage day ended. The speaker also highlighted a more specific attendance figure: the conference was “mostly proud” to have more than 330 LPs present. These numbers were presented as markers of the audience composition and participation during the conference, not as broader claims about the market or investor landscape. The next point addressed the event’s online streaming experiment. The speaker said that, for the first time this year, the conference had tried online streaming, and that several people had joined online through a virtual pass. They also explained that recordings were available in the Dataderm VC community. The practical instruction was straightforward: attendees who had not yet checked it out were encouraged to check the newsletter and subscribe in order to get access to the content. This was the only specific follow-up action attached to the digital offering. The closing ended with immediate logistics and an informal invitation. The speaker thanked the audience again, said people could head over to Steven Bartlett if interested, and noted that that session would start in about thirty minutes. They then looked ahead to the evening by saying they would see attendees “tonight for some liquid networking,” before closing with “Thank you, and ciao.” The tone throughout was brief, appreciative, and upbeat, focused on attendance, access, and next steps rather than new strategic content.
“We had more than 2,000 investors here at the conference and everyone is still around.”