Farewell to the Old Project, Boldly Venture into Silicon Valley: From Never Pitching to a Week of Extreme Fundraising that Triggered VC Bidding

This episode is worth studying more than the surface-level "raising $2.4 million from a16z in a week." Because it truly discusses three things: First, why a team that has already generated revenue should still abandon an old project. Second, why a product that seems like a "cloud drive" may further evolve into the data infrastructure for AI Agents. Third, that Silicon Valley fundraising is not really about pitch language, but about product momentum, time compression, signal creation, and choice. However, there are a few points in this material that need calibration. After calibration, the value of the Space case actually increases. 1. First, clarify a few key facts: $2.4 million in funding is real, and it is indeed Pre-Seed. Space Computer announced on August 18, 2026, the completion of $2.4 million Pre-Seed funding, led by a16z Speedrun, with Golden Ventures, Northside Ventures, and about a dozen angel investors from companies like Parsec, Sentry, Stan, Superwhisper, Modem participating. The three founders are: Matthew Ao, Arihant Bapna, Jason Zhao. Additionally, note: "a16z led $2.4 million" does not mean a16z invested $2.4 million themselves. The official a16z Speedrun currently states that the maximum investment from a single company is about $1 million. The $2.4 million is the total amount for the entire funding round. This is a common conceptual error in financial reporting: Round Size ≠ Lead Investor Check Size. 2. The phrase "completed funding in a week" is suggested to be written as the founders' review perspective. Public information can confirm: They went to Founders Inc. in San Francisco during the fundraising process, later entered a16z Speedrun, and completed this round of funding. But public press releases do not confirm day by day: Meeting Jonathan Lai on Monday, which day received the term sheet, which day Golden joined, and exactly which day closing occurred. So the formal course is best written: "According to the team's funding review in the video, the key funding phase was rapidly completed in about a week." Do not turn the timeline in the video into independently audited facts. 3. The biggest historical correction in your material is actually Eden / Kortex. Matthew has publicly written a very worthwhile review: "Nine Lessons from $3 Million Mistakes." It reveals: Their initial project in 2023 was Kortex, an AI note-taking/Second Brain product for content creators. Over two years, they spent about: $3 million in operating costs. The team once reached: About $60,000 MRR. This is not a completely unused product. The real problem is: Stagnant growth + high churn + no strong switching reason for the product + extremely complex technical architecture. Later, Matthew and Ari secretly rewrote the product in Japan for three weeks, ultimately abandoning Kortex and evolving it into Eden. Matthew still writes in his GitHub profile: Eden once reached about $50K MRR / $600K ARR / 25K MAU. These are self-reported data from the founders. So this is not: "Failed project → Space." The truly advanced version is: "A project with revenue but insufficient PMF → proactively abandoned → seeking larger technical problems." This is completely different. 4. This is a cruel rule that entrepreneurs must know: Revenue does not equal Product-Market Fit. A startup: $60K MRR. Annualized: $720K. Many people might think: "Why stop if it's making so much money?" Because revenue is just a signal. What really matters is: Retention. Churn. Organic Growth. CAC. Expansion Revenue. Whether users are dependent on the product. Whether customers actively recommend it. If every month: A large number of new users come in, And a large number of old users leave, What you might have is: Leaky Bucket. A bucket with holes. Constantly pouring marketing expenses in, But the bottom of the bucket keeps leaking. This is why: MRR cannot independently prove PMF. 5. The most valuable lesson from Matthew's last company is that it explains why Space is interesting. Kortex made many typical "smart engineer mistakes." For example: Building Authentication. Building Billing. Complex Infrastructure. Designing architecture for large-scale users in advance. Even messing with Keycloak. Matthew later estimated that unnecessary infrastructure work alone consumed hundreds of thousands or more in engineering costs. In the end, he summarized a good principle: For things that are not core differentiators, try to buy; only build what truly belongs to the core product. This principle is particularly interesting when applied to Space. Authentication: Should not be built. Billing: Should not be built. But: The filesystem itself is the product of Space. This wheel must be rebuilt. This is the judgment that mature founders should have regarding: Build vs Buy. Not: "Do everything yourself." Nor: "Call APIs for everything." But rather: Concentrate the best engineering time on what constitutes the true moat. 6. What exactly is Space? It is not just "an infinite capacity cloud drive." This is the most important technical concept of this episode. The traditional Dropbox / Google Drive approach is roughly: Cloud ↓ Sync ↓ Local File ↓ Application. In other words: When you process an 80GB file in Premiere, You usually need to first: Download, Sync, Or at least cache a very large portion. What Space wants to change is this path. It allows the system to see a filesystem namespace similar to a regular hard drive. Premiere, DaVinci, Photoshop, Blender see regular files. The actual data exists in the cloud. When an application requests a certain segment of data: Space only sends the required byte range. The Space website now clearly states that it only streams the byte ranges that applications actually request and only caches the portions that users actually access. So the core is not: "A larger cloud drive." But rather: Decoupling Data Location and Data Usage. 7. "Zero Bytes on Disk" also requires a technical understanding, not to be taken literally. The website marketing language is: Zero bytes on disk. But the website also states: Caches only what you touch. This means that during operation, there may still be: Temporary Cache. Buffer. OS Cache. Application Cache. Otherwise, many workloads cannot run at high performance. Therefore, a more professional understanding should be: No need to save a complete permanent local copy. Not: Physically not using a byte of RAM/SSD. This distinction is important. 8. "Infinite Storage" is also a product abstraction, not infinite free capacity. Space's current homepage states: Infinite space on your computer. But it has now announced prices: Individual: $15/month, including 1TB. Teams: $30/member/month, each member contributes 1TB pooled storage. More capacity requires additional payment. So: Infinite ≠ Free unlimited cloud storage. The real meaning is: Your computer's local disk capacity no longer determines how much data space you can access. A 256GB MacBook can have: 20TB, 100TB, 1PB of logical file space. This is called: Storage Virtualization. 9. The statement in your material that "the network is already several orders of magnitude faster than mechanical hard drives" is something I suggest modifying. This is technically overstated. 1Gbps network theoretical peak is about: 125MB/s. Ordinary mechanical HDD sequential reads can usually achieve: About 100—250MB/s. So Gigabit networks are not several orders of magnitude faster than HDDs. 10Gbps networks theoretically: About 1.25GB/s, Can exceed mechanical hard drives. But modern NVMe SSDs can achieve: Several GB/s. And the internet has a particularly large problem: Latency. Panzura even emphasizes when discussing global cloud file systems that the real trouble with remote access is WAN latency, not just bandwidth. So the magic of Space is not: "The internet is faster than hard drives." But rather: You don't need to transfer the entire file. This is a completely different logic. 10. For a 100GB file, what you actually need may only be 500MB of it. For example, video editing. You have: 100GB of raw material. But during the editing process, At some minute you might only be: Previewing a few clips, Scrubbing certain positions, Reading data near keyframes. If the system is smart enough: 100GB Does not mean you must first transfer: 100GB. But rather: Fetch What You Touch. This is similar to Netflix's logic. When you watch a two-hour movie: Netflix does not wait for the entire movie to download before letting you play. 11. However, video editing is much more difficult than Netflix. Because Netflix is: Sequential Read. Reading from front to back. Editors may suddenly jump from: 00:02:31 To: 01:17:42. Then drag back to: 00:26:04. This belongs to: Random Access Workload. So the real difficulty lies in: Prefetch. Caching. Byte-range prediction. Metadata. Latency hiding. Codec behavior. Concurrency. File locking. This is where Space's true engineering value lies. 12. And I want to particularly remind: Space is not the first to think of this. This is a place where we must remain clear in analyzing this project. LucidLink's core capability currently being publicly sold is already: "Using cloud files like local disks." It also: Only streams the data required by applications, Supports Premiere, Final Cut, DaVinci, CAD, and other large file workflows, And publicly claims to have over 100,000 users. In the enterprise market, there are also: Nasuni. Panzura. They are also doing: Global Filesystem. Edge Cache. Unified Namespace. Cloud Object Storage. Therefore: Space did not invent the category of "cloud streaming file systems." This must be made clear. 13. So why did a16z still invest? This is a good question. Because investors are never just buying: "Has anyone done this feature before?" Investors will ask: What new changes in the market make it possible for a new winner to emerge now? Space's answer is actually: In the past: Filesystem services: Human. In the future: Filesystem serving: Human + AI Agent. This is its truly non-consensus bet. Jonathan Lai's public explanation of the investment also clearly emphasizes: Space aims to provide both human creators and AI agents with the same primitive: Instant access to the required data. 14. Why do AI Agents make "file systems" important again? Today, many AI Agent data workflows are: Company files ↓ Uploaded to the system ↓ Parsed ↓ Chunked ↓ Embedded ↓ Indexed ↓ Vector Database ↓ Agent queries. This creates very complex data duplication. Each AI product: Builds its own index. Does ingestion again. Stores a context again. Space's long-term thesis is: Why can't agents, like real employees, directly see the same company file system? If made into: Agent ↓ Filesystem ↓ Which bytes/files/folders are needed ↓ Direct access. Data infrastructure would be much simpler. 15. This is actually competing for a very large position: Agent Data Layer. In the future, agents will need not only: LLM. But also: Compute. Memory. Permissions. Files. Apps. Identity. Tool Access. Among them: Files Are almost the common primitive of all knowledge work. Word documents. Videos. PDFs. Code. CAD. Images. Training data. Datasets. If Space ultimately controls the access layer between AI Agents and unstructured data, Its TAM will no longer be: "Selling cloud drives to YouTubers." 16. So video creators are actually just a wedge. There is a particularly important concept in entrepreneurship: Wedge Market. You don't need to solve the whole world on the first day. First find: The strongest pain points, The most willing to pay, The easiest to showcase the product's magic group of people. For Space: Video Editors are very attractive. Because: A 20GB file, Ordinary people cannot see the cloud drive problem. A team processing: 3TB RAW footage daily, The problem becomes extremely obvious. 17. Then it can expand from Video to AEC. AEC: Architecture / Engineering / Construction. Architects may handle: Revit. CAD. BIM. Large 3D models. These files are also: Huge. Cross-office. Multi-person collaboration. Space has already clearly listed Video, Marketing, AEC as the first industries to enter, and later hopes to expand to AI training data, computer vision, enterprise data systems, world-model pipelines. This is: Beachhead → Horizontal Infrastructure. First vertically penetrate, Then expand horizontally. 18. But the biggest technical risk for Space is actually here. Expanding from: Streaming video in Premiere To: All files Is extremely difficult. Different applications have completely different: Access Patterns. Locking Behavior. Read/Write Patterns. Metadata Requirements. Cache Requirements. For example: Video: A lot of sequential + seek. Code: A lot of small files. CAD: Complex random read/write. AI training: Extremely high throughput sequential datasets. So: "Supporting all files" is a very large engineering commitment. This is also why file systems are one of the hardest things to get right in computer science. 19. The second risk: Consistency. Assuming: Editor A in New York modifies a 60GB file. Editor B in Los Angeles opens it simultaneously. AI Agent C is also modifying metadata. Who has the latest version? What if B modifies an old version? What happens after disconnection? How to merge after reconnecting? This involves: Consistency Model. Locking. Conflict Resolution. Versioning. Atomic Write. Recovery. These are the real questions enterprise customers will ask. Not: "Does the demo look cool?" 20. The third risk: Security. If Space ultimately becomes: A data layer used by everyone + all agents, Then what it controls is very sensitive. In the future, it must handle: RBAC. SSO. SAML. Encryption. Audit Logs. Data Residency. Agent Permissions. Least Privilege. Revocation. Ransomware Recovery. The current official enterprise plan has already started listing: SSO/SAML, auditing, private cloud, on-prem capabilities. This indicates that the team knows: The real money is likely in Enterprise. 21. The fourth risk: Cloud storage unit economics model. This is a layer that many people overlook when looking at SaaS. Space currently Individual: $15/month including 1TB. Taking Cloudflare R2 as a pure industry cost reference—note, I am not saying Space necessarily uses R2—R2 Standard's current public price is about: $0.015/GB/month. That is, 1TB itself is about the $15/month level of public storage price. So you will immediately find: If users really fill: 1TB full, And your underlying costs are close to public cloud prices, Just relying on the $15 subscription: Gross margin may be very difficult. 22. Therefore, what Space really needs to manage is not "storage capacity," but Storage Economics. It needs to utilize: Average users will not use up their quota. Volume Discounts. Cheaper Object Storage. Caching. Deduplication. Cold Storage Tiering. BYO Bucket. Enterprise Pricing. Seats. Value-added features. Even in the future: Compute / Agent Services. To establish a healthy Gross Margin. This is also why mature cloud storage companies' Unit Economics are very complex. Selling Storage can easily become a Commodity. What is truly valuable is: Access + Workflow + Collaboration + Intelligence. 23. This is also why I believe Space must not position itself as "a better Dropbox." If it is just: Dropbox + Streaming, The giants can completely replicate it. A more valuable positioning should be: The Filesystem for the Agentic Computer. Then continuously build up: Semantic Search. Agent Sandbox. Version History. Permission Layer. Workflow. Compute. If it reaches this level, Space will not be selling: TB. But rather: Productivity. 24. This aligns with Matthew's latest public vision. He publicly describes the long-term state of Space Computer as: In the future, you can: Start dozens of agents, Close the laptop, Leave. Agents in the cloud: Edit videos. Process CAD. Write code. Build datasets. When you come back, The results automatically appear on your computer. In other words: The laptop is no longer the computer. The laptop is just: An interface. The real computer is in the cloud. This is actually much bigger than "infinite storage." 25. This may be the most ambitious layer of Space: "Thin Client" is back. The history of computers has always been a pendulum. Large Mainframe: Centralized computing. ↓ PC: Localized computing. ↓ Cloud: Server centralized. ↓ Smartphone: Local + Cloud. ↓ AI Agent: May again massively return to Cloud Compute. If: Storage is in the Cloud. Compute is in the Cloud. Agent is in the Cloud. Memory is in the Cloud. Then what does a future laptop need? Mainly: Display. Keyboard. Camera. Network. Local Cache. Secure Enclave. Thus: Physical Computer → Window into Compute. This is what they call: Infinite Computer. 26. Looking back at the fundraising: their poor pitch in the first two weeks is actually very normal. Entrepreneurs often think: Fundraising is about telling stories. In reality, VCs will continuously ask: How much are you raising? Why raise now? Why this amount? How long can you survive? What milestones will you achieve? What do you need to prove before the next round? Who is leading? How big is the market? Why now? Why you? Who are the competitors? What is your GTM? By the way, correct: The "GTM (Go-To-Market Strategy)" in your material Is incorrect. GTM = Go-To-Market. In Chinese, it should be called: Market entry/customer acquisition and commercialization strategy. Not "decentralized market." 27. Why is answering "maybe $3 to $5 million" so bad? Because what investors hear is not: "This founder is flexible." But rather: This founder does not know how much they need. A professional answer should be similar to: We are raising: $2.5M. Planning to provide: 18—24 months runway. Funds will primarily be used for: Core filesystem engineering, Windows, Security, GTM. Before the next round, we hope to achieve: X paid customers, Y ARR, Z retention. This is called: Financing Plan. The fundraising amount is not arbitrary. But rather derived from: Milestone Backtracking: Burn Rate Then backtracking: Capital Need. 28. The true essence of fundraising is "selling future milestones to the capital market in advance." VCs invest in companies that do not have mature cash flow today. They are buying: A probability distribution at some point in the future. So founders are actually saying: Give me $2.4M, and I can advance this technical hypothesis to the next value inflection point. This inflection point could be: 10,000 paid users. Forbes reported that one of Space's initial goals is to reach 10,000 paying users. Doing a pure mathematical thought experiment based on the current cheapest $15/month Individual plan: 10,000 × $15 Is: About $150K MRR / $1.8M annual revenue scale. Of course, actual revenue will be affected by: Teams, Enterprise, Trial, Churn, Annual Plans, Storage Overage. But you can immediately understand: Why 10K paying users is a meaningful next round milestone. 29. Why is "Build in Public" particularly effective for fundraising? Because investors fear one thing: Information coming entirely from the founder's mouth. Pitch Deck: You wrote it yourself. Market Size: You calculated it yourself. Roadmap: You drew it yourself. But: Weekly releasing demos. User comments. Users actively sharing. Waiting list growth. Founder audience growth. Customer inquiries. These are: External Signals. The more signals, The less investors need to rely entirely on the founder's narrative. 30. And Jason is a severely underestimated asset in this team: Distribution Founder. The most common structure in tech startups: Two engineers. One product. No one knows how to pitch. Space is different. Jason has long-term YouTube content experience. The team disclosed that they currently have over 80,000 organic audience members, and at the time of the fundraising announcement, about 100 users/teams were in the testing phase. This means: Engineering And Distribution Have been in the same founder team from Day 1. This is very valuable. 31. YouTube is not their "marketing department," but an asset. Why? Traditional startups: First fundraise. ↓ Develop product. ↓ Hire Head of Marketing. ↓ Buy ads. Space: The founder already possesses: Camera Skills. Editing. Storytelling. Audience. Content System. So Customer Acquisition Cost may be lower from the start. This is called: Embedded Distribution. If two companies have identical product capabilities, One company spends: $300 per user acquired. The other: $30. Five years later, the results may be completely different. 32. The so-called creation of investor FOMO is not essentially about "lying that others are investing." The truly healthy FOMO comes from: Product Momentum. Investor Momentum. Customer Momentum. Round Scarcity. Meeting Compression. For example: Monday: Investor A interested. Tuesday: Investor B second meeting. Wednesday: Lead discussion. So Investor C knows: If they wait two weeks, The round may end. This creates: Time Pressure. 33. Absolutely do not learn to create "fictional FOMO." Do not say: "We already have a term sheet." When in fact, you do not. Do not say: "a16z wants to invest." If it was just a one-time conversation. This involves: Reputation. It may even involve legal issues. What is truly valuable in Silicon Valley is: Reputation Capital. You can deceive one person in a funding round, But your entire career may still have: 30 years. No need for that. 34. The detail about Golden Ventures adjusting their schedule the next day is very worth learning. In investors' mouths: "Interesting." Has almost no informational value. What you should really observe is: Behavior. How long do they follow up? How long until they arrange a second meeting? Do they proactively pull in partners? Do they ask for customer references? Do they inquire about round allocation? Do they request a data room? Do they discuss terms? Real interest: Will compress time. So during fundraising, do not just listen to: Words. Watch: Calendar. 35. The essence of Warm Introduction is not "relying on connections." Many people misunderstand: Silicon Valley relies on connections. More accurately: Referrals reduce Information Asymmetry. Assuming a VC receives: 10,000 pitches a year. They cannot research them all. If a founder they trust says: "These three young people are particularly strong, you must meet them." This is: Signal. Equivalent to: Reducing the investor's search cost. So true networking is not: Adding 5,000 LinkedIn connections. But rather: Someone is willing to back you with their reputation. This is completely different. 36. The fundraising for Space also has a very nice location effect: San Francisco. Many people will ask: Why go to San Francisco in the online era? Because fundraising and entrepreneurship involve: Network Density. In one day, you can: Meet a founder in the morning. Encounter an investor at noon. Demo in the afternoon. Dinner in the evening. The next day, someone introduces you. The speed of information dissemination is extremely fast. This is the clustering effect. And: Wall Street. Hollywood. Shenzhen hardware. London financial district. Geography is not magic. The real value is: Network Latency is extremely low. 37. However, a successful fundraising should never be understood as a successful startup. This is the easiest place for young entrepreneurs to misunderstand in such videos. Space: Raised $2.4 million. Does not equal: Earned $2.4 million. Even less equals: Founder's worth increased by $2.4 million. Fundraising means: The company sold part of its future equity, In exchange for: Runway. $2.4 million is essentially: Shareholder capital that must be effectively allocated. From the moment the money is received: The timer starts counting down. 38. Moreover, the current valuation and specific dilution ratio of Space have not been disclosed. So do not infer: $2.4 million round To estimate: Company valuation of $10 million, $20 million, $30 million. There is no basis for that. Pre-Seed often uses: SAFE. It may also use: Priced Equity. There may be: Valuation Cap. Discount. MFN. Pro Rata Rights. These are currently insufficiently disclosed in public information. The correct statement is: Public information is limited, and confirmation is currently not possible. 39. The biggest competitive risk for Space today is actually not another startup. But rather: Apple. Google. Dropbox. Microsoft. Amazon. And players like LucidLink that already have mature technology and customers. Why? Because Space is located: Between OS And Cloud. Above are: Applications. Below is: Object Storage. This position is very valuable, But also easily squeezed by upstream and downstream. 40. Therefore, Space must ultimately form the value of a "neutral layer." If in the future: Your files may be in: AWS. Google Cloud. R2. Azure. Private Cloud. NAS. Different agents also run on: OpenAI. Anthropic. Google. Local models. If Space can become: Vendor-Neutral Data Plane, It may generate real strategic value. Because: Apple will favor Apple. Google will favor Google. AWS will favor AWS. But an independent company can say: I connect to everyone. This is similar to: Stripe does not need to own a bank. Cloudflare does not need to own all source servers. Snowflake does not need to own all data sources. A neutral layer is sometimes the biggest opportunity. 41. There is also a bigger opportunity: Filesystems may once again become the "operating system interface" of the AI era. The past GUI abstractions were: Folder. File. Desktop. In the future, agents may not even need a Folder UI. They just need to ask: Find all video materials related to Nike from the past three years, pick out all the street shots from New York, and make a 30-second advertisement. Spacebar / Semantic Search ↓ Locate data ↓ Agent fetches Byte Ranges ↓ Video Agent edits ↓ Results write back to Filesystem. At this point: Filesystem Upgrades from: "A place to store files" To: "Agent Execution Substrate." This story starts to become very large. 42. But do not underestimate the last 10% of difficulty. Getting the demo to: 90%. Is easy to impress people. Achieving: 99.999% In an enterprise-level file system is the real death zone. Once: File corruption. Sync loss. Version overwrite. Permission leakage. Could directly lose an entire studio client. So the ultimate moat of Space may not only be: Streaming Algorithm. But also include: Reliability Reputation. Companies like Dropbox, AWS, Microsoft have one of their core assets as: Users trust that files will not disappear. This trust takes many years to build. 43. I believe the first thing these three founders did right is that they solved their extremely painful problem. Jason has been shooting videos for years, Accumulating dozens of TB of material. The previous company also moved a large amount of video daily due to content production. This problem is not: What VC reports told them. It is: A problem they complain about every day. The official funding explanation from Space also directly cites this experience as the product's origin. This is called: Founder-Market Fit. You possess pain knowledge that others do not have. Workflow knowledge. Customer language. 44. The second thing they did right: solve specific problems first, then tell a billion-dollar narrative. Initially: "My hard drive is not enough." Then: "Why can't cloud files behave like local files?" Then: "Infinite Storage." Then: "Filesystem for Humans + Agents." Finally: "Infinite Computer." This is a very beautiful: Vision Expansion. Good startups are often: Very specific problems, Very huge visions. 45. The third thing they did right: the product demo is instantly understandable. VCs love a product that can: Create a Magic Moment in 30 seconds. 256GB Mac. ↓ Open 20GB BRAW. ↓ Drag directly into Premiere. ↓ Play normally. ↓ Finder shows 0 bytes. No need to explain: 30-page deck. Users: "Wow." This is: Demoability. Extremely valuable for Pre-Seed fundraising. 46. The fourth thing they did right: they treated the fundraising process as product iteration. Early investors asked: Why Now? Did not answer well. Went home to fix it. Asked: Market Size? Fixed it. Asked: GTM? Fixed it. Tested again in the next meeting. So the Pitch Deck is essentially: A Product. VCs are the users. Objections are feedback. Pitch is iteration. Conversion is term sheet. 47. This is completely the same as sales. The worst founders: Get rejected 10 times, And conclude: "VCs don't understand me." Excellent founders: Get rejected 10 times, And discover: Everyone is confused by page 7. Then: Delete it. This is: Feedback Velocity. And it is exactly the same as product entrepreneurship. 48. If I were to judge the investment logic of Space, I would break it down into four layers. The first layer: An obvious real pain point. Moving large files is indeed painful. The second layer: Technical Timing. High-speed networks, cheap object storage, modern filesystem APIs make the experience feasible. The third layer: New demand from AI Agents. The amount of data generated and read by machines may far exceed that of humans. The fourth layer: The option value of expanding from Storage Product to Compute Platform. What VCs are really buying is primarily: The fourth layer. Because the first layer may only be one company. If the fourth layer is valid: It could be a platform. 49. But I would not currently call Space "already disrupting local storage." Too early. Currently, a more accurate statement is: Space is attempting to redefine how cloud data is used by humans and AI Agents in the manner of local file systems. Because: User scale is still small. Revenue has not been disclosed. Retention has not been disclosed. Storage Gross Margin has not been disclosed. Enterprise Adoption has not been proven. Competitors already exist. This is a: Very interesting Pre-Seed Bet. Not yet a validated winner. 50. The ten things this video should truly teach entrepreneurs: First, projects with revenue can also lack true PMF. Second, the greater the sunk cost, the more ability is needed to recalculate the future. Third, for infrastructure that does not constitute a moat, try to buy; core capabilities must be mastered in-house. Fourth, the best entrepreneurial topics often come from problems that founders repeatedly encounter at a high cost. Fifth, wedges should be narrow, visions can be huge. Sixth, fundraising is not a storytelling competition, but a comprehensive pricing of product, team, market, timing, and momentum. Seventh, true FOMO comes from real momentum, not from rhetoric. Eighth, Build in Public is not just marketing; it can also create customer signals and investment signals. Ninth, infrastructure entrepreneurship ultimately competes not on demos but on Reliability, Security, and Unit Economics. Tenth, one of the biggest opportunities in the AI era may not be to recreate an agent, but to build the foundational layer that all agents must rely on. 51. I will ultimately elevate the entire episode into a formula. The story of Space is not: Having an idea → Going to San Francisco → Meeting a16z → Raising $2.4 million. The real chain is: Failures from the old project ↓ Discovering structural pain points ↓ Founder-Market Fit ↓ Minimal Prototype ↓ Visual Magic Moment ↓ Self-owned Distribution ↓ Build in Public ↓ User Validation ↓ Fundraising Momentum ↓ Capital Inflow ↓ Expanding from a single Use Case to an Infrastructure Thesis. This is what ordinary entrepreneurs should replicate. 52. I suggest upgrading the title. Among your four titles, I think the second direction is the best, but "cloud-based streaming editing" makes Space sound smaller. What it really wants to do is no longer just an editing tool. I recommend: "Not a Cloud Drive, but the Data Access Layer of the AI Era: How Space Pivoted to Secure a $2.4 Million Investment from a16z" If more focused on entrepreneurship: "From $3 Million Lessons to a16z Investment: Space's Pivot, Fundraising, and 'Infinite Computer' Ambition" If more focused on technical trends: "Reinventing the Filesystem: How Space Enables Humans and AI Agents to Share an 'Infinite Computer'" If more focused on communication: "Three People, One Week, $2.4 Million: Why They Made a16z Bet on 'Eliminating Hard Drives'?" If included in formal courses, I recommend the second one. Because it highlights the most precious layer: What is truly impressive about this story is not "raising $2.4 million in a week," but that they first spent $3 million learning what not to do, and then compressed these costly failure insights into the next company. The true compounding in entrepreneurship is not just the compounding of money. There is a more scarce type: The compounding of experience after correcting erroneous understandings. And the core worth continuing to observe about Space is not whether it can reduce a few GB downloads for Premiere, but whether it can ultimately cross from "cloud file systems" to become the "universal data layer for AI Agents." If it cannot cross over, it may just be a decent storage SaaS; if it can, the story will truly start to grow.