What Do Products Look Like in a New Era When Software Is Easy to Create? | A Conversation with Albert

Summary

• Albert sees software’s value chain potentially taking on a barbell shape: a small number of model companies capture most of the value, 3-5 person OPCs absorb fragmented demand, and mid-sized software companies face the greatest risk. Model capabilities may erase existing software moats; startups that only patch edge-case functionality are “picking up pennies in front of a steamroller”—they may grab some if they move fast enough, but the pennies are still pennies. With coding agents, 3-5 people may achieve the output of teams that once needed 30-50.
• Making software easy to create does not make software companies easy to run. Only 5-6 people on Albert’s team actually code; they built dozens of products in a few months, but the postmortem was “the returns were zero.” Internal tools can be usable overnight, but products aimed at the market still have to solve operations, growth, and standardization—and may not offer enough depth of value.
• Long-tail products are more likely to win through experience, taste, and emotional value than through broadly applicable functional demand. A vertical-layout classical Chinese note-taking app, a world-street-view explorer, or a poetry-and-geography map is unlikely to become a conventional large business, but can give users a sense of identification. Albert’s shorthand is that software will “look more and more like Pop Mart, and more and more like the cultural industries.”
• The next opportunity is not only to build creation tools, but to connect new groups of makers and generate “resonance.” Model companies control the publishing layer, but may not control the consumer containers—web, mobile, or desktop. Albert’s team is therefore building a community product with about 8 people involved; the immediate goal is not monetization, but helping creators answer the question: “Whose recognition matters to me?”
• Qu Kai remains cautious on the coming supply explosion: AI may push an existing experience from 80 to 85 or 90, but may not deliver a step-change from 20 to 70. Albert agrees that concentration at the top and distribution patterns will persist, but believes new media capabilities will surface new creators, just as short-video creators were not simply a continuation of early DV videographers. He also concedes that the change may be less dramatic in usage value alone; emotional value may be more visible.
• If mid-sized software companies shrink, private markets may shift from betting on scale to revenue sharing, cash flow, and small cultural products. Albert expects many small teams with a few million dollars in revenue; investors may fund them like a shop and take a share of the proceeds. Organizations may also move from a business model to an impact model—create impact first, then convert it into economic returns.
• The shortest path to AI monetization may not be building a product, but letting tokens generate trading returns directly. Albert proposes the One Person Fund: individuals use AI to build prediction-market or crypto strategies, then raise capital, sell the strategies, or take bets on a platform. Polymarket’s information processing and arbitrage opportunities may already offer a glimpse of what is possible: “the answer is already in the question”（谜底已经在谜面上了）.

Deep dive

1. Model Upgrades Put Founders on a Permanent Pivot Treadmill

• Qu Kai sees the same pattern across founders: they are excited about an opportunity one month, then discover the model has absorbed its value the next and are forced to pivot again. OpenClaw was hot a few months ago; now there are reports of Hermes and the next-generation OpenClaw, making it difficult for him to get excited about another application-layer wave.
• Albert acknowledges that the pessimism has a first-principles basis. Model companies may create the overwhelming majority of the value, while applications that only fill gaps are left “picking up pennies in front of a steamroller”—move fast enough and you may pick some up, but they are still just pennies.
• The counterpoint is that some small teams and super-individuals now have 100x leverage, which is why OPCs can exist. This does not necessarily mean one person; it could be 2-3 or 3-5 people operating with the output of a team that previously required 30-50.

2. Dozens of Internal Products Prove Creation Efficiency—but Not Commercial Value

• Albert says the team felt the pace of technical iteration accelerate after Claude 4.5 launched on November 24, followed by Claude 4.6 in February. The 5-6 people who actually code had built dozens of things within a few months, but the postmortem was “the returns were zero.”
• The issue is not whether they can build, but whom they are serving. It is easy to make tools for the team’s own document-versioning habits, workflows, and small annoyances. Once the product faces external users, the full set of company problems returns: operations, growth, compatibility with user habits, and standardization.
• The internal products do get used, but they are highly personalized. Tool enthusiasts will build one product for every pain point; everyone else may reuse a colleague’s work or choose a mature product. Albert thinks demand for forks on the individual side may be limited. Qu Kai says it is too early to draw that conclusion; the answer depends on the product type, how it is discovered, and what users want it for.

3. Infrastructure Compresses Development Time, While Conversation Logs Expose Human Scarcity

• The team brought its former stack—Excel plus Supabase, backend deployment, databases, and related steps—into an integrated internal environment. With that infrastructure in place, Albert came up with an idea at 9 p.m. half a month ago, had finished coding by 3 a.m., and was already using it.
• They also built an experiment that reads local coding-agent conversations, scores them, and generates a personal page. Albert’s logic is that the agent itself comes from the model company’s capabilities; the real source of differentiation is how a person frames requests and what they continue discussing with the agent.
• The most revealing module is called “What do you reject?” In 99% of cases, users have little to teach the agent; non-engineers in particular may lack the ability to evaluate the solutions it produces. Rejection still exposes taste: some people obsess over architecture, others care about UI, and others insist on product definition. Since the project launched, Albert has already seen about 10 similar products in the market.

4. Software’s Value Chain Will Consist of Model Companies and Micro-Teams

• The internet and mobile internet moved information, data, and industry workflows onto computing devices, creating a large number of jobs in the process. Albert’s extrapolation is that every task performed on a computing device could gradually become “a person directing a coding agent to do it,” especially in productivity work.
• Models are heavyweight systems that only a limited number of participants can build. Codex and Claude Code are, in essence, tool containers for model companies. If natural language is enough to invoke these capabilities, Photoshop, IDEs, and large swaths of vertical productivity software may lose their reason to exist independently.
• The result is a barbell structure: model companies control most of the value at the top, while the bottom consists of highly fragmented small teams. Mid-sized companies whose moats were built on accumulated software assets and organizational costs become the most awkward part of the stack.

5. Long-Tail Software Will Shift from Functional Competition to Taste Competition

• Albert does not fully accept the hardware analogy that the future will have only 10 software brands. Productivity entry points may consolidate, but everyday small products could become more fragmented because extremely low costs make it possible to serve even a resonance group of just 10,000 people.
• His sharpest example is a colleague’s vertical-layout classical Chinese note-taking app: “Who needs this?” The answer is that its creator loves it intensely. The design and color grading of imaging products such as Dazz can also create a differentiated experience and a loyal fan base. Style and emotion beyond function may give software a premium rarely seen in traditional industries.
• This is closer to art, meme coins, or cultural consumption. Outsiders see something and ask, “This works?” while users identify with its meaning and taste. Albert therefore calls the products of the future “experience products,” rather than merely content or tools.

6. The Challenge for New Platforms Is Defining the Consumer Container, Not Copying Creation Tools

• Albert distinguishes the publisher from the container. Model companies control the former because the creation tools sit with them; the latter could be an operating system, mobile software, a mobile app, a website, or PC software—and may not be controlled by the model companies.
• A startup that builds both the creation tool and the publishing layer still has to fight coding agents from model companies on price, unless it subsidizes the product. The more viable path may be to identify a new content format first, then design a native container for it. Albert also thinks the container may not be necessary at the current stage, because it can constrain what creators are able to make.
• Some people currently publish skills on GitHub, while others post small products on Xiaohongshu. It is unclear whether those legacy platforms will continue to absorb the output. If a new creator cohort develops its own culture, Albert thinks existing platforms may be incompatible with how its work is presented, distributed, and connected.

7. Supply Will Explode, but Whether Consumer Value Changes Qualitatively Remains Contested

• Qu Kai’s supply-side rebuttal is concrete: even if a Douyin user watches 1,000 videos a day, the platform already has tens of thousands or more pieces of content. Adding AI video does not relieve any scarcity in the short term, and most long-tail apps in the App Store have never been used.
• Albert separates the platform from the consumer. The platform may already have an ample pool of quality content, but greater variety can still retain people who previously could not find what they wanted. As controllability and consistency improve, imaginative creators may push the medium into new forms.
• Qu Kai sees this as moving from 80 to 85 or 90, not jumping from 20 or 30 to 70 or 80. Albert agrees that concentration at the top and distribution patterns are inevitable, but emphasizes that new tools will surface people who master entirely new techniques. He later concedes that the change may not be very large in usage value alone; emotional value may be more apparent.

8. The Real Long Tail Comes from Life, Not Just Jobs

• Albert’s team built a world map where clicking a location links to a local walking video on YouTube. One colleague used it for world exploration; another opened it as a sleep aid while suffering from insomnia. It reminded Albert of an idea he had in 2023: using bedtime-only content and AI chat to help people fall asleep, though that was not the same product.
• Qu Kai offered another product idea: connect classical poems to geographic locations. Click “The waterfall plunges 3,000 feet,” and the corresponding waterfall appears. It would be difficult to commercialize, but could add “a little bit of small pleasure” to daily life. Creators could share their work on Twitter and attract followers who recognize their taste.
• Qu Kai notes that many people are beginning to discuss the problem of sharing skills selflessly without being able to monetize them. Albert sees monetization as a product of costs and organizational objectives: companies have shareholders, employees, and growth obligations. Individuals with almost no costs can live a low-power life, make what they enjoy, and not require every work to generate revenue.

9. What a Community Really Delivers Is “Resonance”

• Albert says the team went through an overcorrection in 2023, moving from pessimism to an obsessive focus on profitability. After becoming profitable, the work felt “not very meaningful,” so the team returned to the question of what the era actually needed. Its answer was makers, maker-designers, and new creators who had been brought much closer to top programmers by AI.
• He divides the creator lifecycle into “response, resonance, and return.” Response means someone engages with you or a number provides feedback; resonance means asking, “Whose recognition matters to me?”; return is making money. The current product is betting on the second layer, because chemical reactions can occur among nodes of the same kind.
• The company has already demonstrated this kind of network internally. When Albert gets stuck on a page, he brings in a designer to restructure it; when a complex backend needs debugging, he calls in someone else. He wants to turn this kind of ad hoc collaboration into an open structure where nodes discover one another, fill gaps, and create together.
• About 8 people are working on the community product; they spent a month building it and iterated repeatedly. The measure is not simply whether all the features are complete, but whether the medium, network structure, and mode of connection can actually create resonance. Albert is currently designing it for people, but does not rule out a future audience of virtual beings with independent identities; that is not part of the current definition of a creator group.

10. The Universal Belongs to Model Companies; Startups Can Only Bet on the Particular

• Qu Kai estimates that this creator cohort may currently number in the tens of millions, then asks whether it can reach hundreds of millions. Albert’s extreme scenario is that most people may not need to work within 10 years. Qu Kai thinks 10 years may be optimistic; Albert responds that sudden shocks could accelerate the process.
• Albert’s strategic conclusion is that universal capabilities for “ordinary people” will ultimately be taken by model companies. Applications that merely build around those capabilities are “finished.” Startups need to define the particular and wait for it to spread through successive waves of impact.
• He is looking not for people who build tools solely for occupational value, but for people who enjoy creating for its own sake. This network will not be equivalent to short video, nor fully equivalent to GitHub: its scale may not be universal, but its influence can be substantial. The community can absorb some consumption, but defining a unified container too early would also restrict the range of things that can be created.

11. Capital May Shift from Betting on Scale to Betting on Cash Flow, Strategies, and Connection

• If mid-sized SaaS companies disappear and small software products become cultural products, the traditional exit logic of private markets will weaken. Albert expects many small teams to reach a few million dollars in revenue; investment may look like funding a shop and taking a share of the proceeds. Software will “look more and more like Pop Mart, and more and more like the cultural industries.”
• The shorter path runs directly from tokens to economic returns. Productivity usually has to monetize through job value, while a trading strategy can ask directly: how many tokens did it consume, and how much money did it make? This leads Albert to imagine VC becoming OPF, or a one-person fund, with individuals using AI to build trading strategies and earn returns in the market.
• Prediction markets such as Polymarket require processing large amounts of information. AI can collect signals, arbitrage weather data, and execute strategies. Albert relays a quant friend’s view that a good strategy might produce several times the return in a year with relative stability, while warning that genuinely intelligent strategies are usually not made public. A crypto platform called Ora—Albert spells it O-R-A—allows individuals to put strategies up for sale, raise capital, or accept bets.
• Returning to the present, Albert does not want to describe the team as “all in OPC.” The platform was designed from Day One around discovery, attention, and connection, and will provide support such as one-click backend deployment and growth. Even if its own form turns out to be wrong, inspiring others to create value would still count as effective resonance.