SPC's Request for Curiosity
Which questions does SPC want to see more of in 2025?
South Park Commons is a community for -1 to 0: the earliest stage of a new endeavor, when you’re searching for ideas and the conviction to pursue them. Searching is the key part. We don’t expect the founders, technologists, and researchers who join us to have a concrete idea of what they want to build. We look for talented people with interesting questions.
The SPC community gives a unique view of what the best founders and technologists are exploring. We put together the list below from that vantage. It covers observations and questions the SPC team developed while working with community members and companies. If you're exploring similar questions, reach out to an SPC team member. We want to meet you.
We don't have requests for startups—we have requests for curiosity.
How will we develop and manage software in a world of 10x to 100x more code?
The world 5 years from now is likely to contain one to two orders of magnitude more code—a large amount of it AI-generated. In that world, what does SDLC (software development life cycle) look like?
- Are current VCS's (version control systems) the right abstraction format?
- Will we still want to do code reviews? What does maintainability look like? How about development planning?
- How will we interleave "code" and "design" in the future? What about other workstreams like CS / support / sales / marketing?
How will we manage an agentic workforce?
In the near future, businesses will likely manage a set of AI agentic employees. You will have agents for customer service, sales, HR, accounting etc. It is likely that a lot of these agents will be provided by different companies.
- How do you coordinate the work between the various agents?
- Is there a shared "system of record" that all agents need to work off?
- Is there a shared connectivity fabric (for knowledge, for communication) that needs to be the underlying mesh?
- How will you oversee this fleet of agents? Does there need to be a global view that manages deployment?
Reach out to: Aditya
Where are the co-pilots and agents for Technical Operations / NOC / SRE work?
We have seen a tremendous amount of activity in the codegen space. Why hasn't there been equivalent interest in Technical Operations or Network Operations Center?
- Are there underlying domain difficulties preventing this?
- What role will co-pilots play for SREs and SWEs in managing current deployments and clusters?
- How will this change with 10x AI generated code?
- How much can you automate vs. assist in this domain?
What will the next generation of foundation models look like?
It's easy to assume that innovation in foundation models is now rate limited on compute and data applied to current architectures. However, we believe there are opportunities to make architectural advances that lead to different configurations of performance vs. compute vs. data requirements.
- What are some core assumptions about current architectures that are worth challenging?
- How do we embed custom knowledge as a core feature of models, as opposed to after the fact techniques like RAG?
- How do we enable interpretability as a core design feature?
Will novel chip architectures be necessary for ubiquitous, "always-on" AI inference?
The widespread adoption of “always-on” AI inference could involve multiple concurrent threads, ranging from real-time object detection (e.g., in AR glasses) to local AI agents performing diverse tasks. This raises critical questions about the hardware and software ecosystem required to enable such capabilities and how much of that will be vertically integrated.
- Are current ARM architectures capable of supporting this use case efficiently, or will new architectures be required to meet performance and power constraints?
- What constraints—such as power consumption, thermal management, or compute bottlenecks—must be addressed/relaxed?
- Is there an updated or restricted version of CUDA that will need to be built?
Reach out to: Ruchi
What is the Drone software stack going to look like?
Drones are poised to play a transformative role in both industrial applications and defense. While there's significant work to be done in nearshoring drone hardware production, our focus lies on the potential structure of the drone software stack.
- Is there potential for a unified, horizontal platform that standardizes core functionalities—such as navigation, communication, and data processing—while enabling customization for specific use cases?
- What specialized verticals could emerge atop such a platform? For instance, AI-powered decision-making, sensor integration, or real-time field analytics.
- How can we lower barriers to entry for software engineers, even those without prior experience in drone technology? Could accessible development kits, robust simulation environments, or open-source frameworks play a role?
What does creative tooling for children look like in the future?
Multimodal LLMs offer an incredible opportunity for children to be creative in a self-directed way. They can create images, videos, stories, cartoons and music. And all of these can be mixed in interesting ways to really delight children.
- How do we create a multimodal canvas that is remarkably simple and intuitive?
- Could there be a modern "Photoshop" for children?
- How might children intuitively use AI in ways that adults wouldn't think of? Do those interactions translate into products for older users?
Why are there no multi-player AI interfaces?
There is a noticeable absence of multi-user AI-native platforms that allow for collaborative content creation and query management, akin to how platforms like Google Docs and Notion enable multi-user collaboration.
- How do we manage shared context and knowledge windows for multiple users while maintaining privacy and without creating confusion or bias towards a particular user’s query?
- How can we enable real-time collaboration with AI assistants, including features like query arbitration and group knowledge sharing?
- How do multi-player AI interfaces manage data access and privacy?
Reach out to: Dylan
What will it take to disrupt low-quality but powerful incumbent forms of credibility?
In certain industries such as corporate auditing or credit ratings, a small subset of companies capture the lion's share of the market—not necessarily because of the quality of the product or even the distribution advantage, but because of brand recognition and reputation/career risk of not choosing them.
- In an environment where an increasing amount of once heterodoxical views are later proven correct, how will that impact people's willingness to accept the status quo?
- As the underlying tools (AI) used to perform work are democratized across incumbents and new entrants alike, will the perceived quality differential persist?
Reach out to: Danh
What kinds of consumer experiences are uniquely unlocked by AI?
Massive consumer outcomes (Meta, Snapchat, TikTok, etc.) have found product-market fit just as novel technologies break through to the ecosystem (mobile, 3G, LTE). If we see "intelligence on tap" as a net-new, publicly accessible technology, what are the novel kinds of networks and applications that will win?
- How will creator-consumer relationships change when mediated by AI? Will the ratio change?
- If (when) users get tired of text interfaces, what are the novel ways users can share data with AI systems? How do novel interaction loops unlock new product types?
- In a world of content abundance, how will consumers discover and personalize what they want?
Reach out to: Jonathan, Gopal, Prateek
What will the creative process look like in the future?
While it may take some time to get there, we're headed toward a future where we will be limited only by our innate creativity, not by our underlying lower-level skills.
- What are the new tools for design and production across software and hardware?
- What will be the new paths toward mastery? Will the underlying skills still be necessary to demonstrate mastery? Or will mastery look completely different?
- What will the new design paradigms be for exploring, creating and iterating on ideas?
- How will we provide pro users the level of control needed to translate ideas with fidelity?
How does content consumption change as production costs move toward zero?
The ZIRP era showed how a massive increase in supply could dramatically change media economics. As the cost for creation drops and more people are able to become creators, how does that change consumption habits?
- How strong are user preferences for shared vs. personalized media? How and where might those preferences shift as hyper-personalization becomes possible?
- As production and consumption become more and more frictionless, will higher-friction experiences or content modalities become more valuable? Where?
- How does the relationship between creators and consumers change?
- Does media consumption become more interactive? Do existing media types start to look more like games?
Reach out to: Jonathan
How (or will) data go from platform owned to user owned?
Users may soon have personalized models they can connect to different tools. Does this happen, and if so, what kind of infrastructure will be needed?
- Will users want to port their data / personal foundation model across services?
- How will user privacy evolve?
- How will businesses adjust to high value users putting a premium on their personal data?
- What role does crypto play in this world?
What is the future of marketing and advertising in an agentic world?
When every consumer has their own agent doing research, shopping and doing other tasks on their behalf, the relationship between advertisers and customers may fundamentally change.
- Will advertising shift to pushed content based on agent activity? Are there completely novel forms of marketing that will be agent-first?
- How will brand experiences evolve when agents mediate consumer relationships?
- How will marketing metrics of CAC, LTV and retention evolve?
- What impact will this have on performance marketing organizations?
Reach out to: Arian, Prateek, Jonathan
How will action-oriented agentic systems evolve when not limited by human computer interaction?
The first wave of companies building agentic workflows are browser automation tooling based on screen recordings of users. As input modalities expand and agents need increasingly less direction from users, we expect the interaction environments to change.
- What if we were to rethink systems from scratch? How many assumptions about digital environments become obsolete with agentic systems?
- What does a website look like that is readable by a computer vs. human?
- How might websites evolve to be essentially boiled down to read/writing directly to databases?
- Where else have we baked human limitations into digital interactions that are no longer relevant?
Reach out to: Arian
What will APIs look like in an agentic world?
Today, APIs are the communication layer of software-to-software interaction. In a future where AI agents make up the majority of "software", we will need to reimagine what an API looks like.
- How will API architectures adapt to agent needs and capabilities?
- How will security protocols and authentication evolve?
- How will AI agents function as APIs to the physical world?
- Will we see a universal "language" or protocol for cross-platform agent interaction?
Reach out to: Finn
What happens as interaction between humans and computers becomes increasingly high-throughput and low-latency?
New models will open up entirely new ‘Human APIs’ that enable us to interact with underlying functions of data and technology more naturally.
- How do human-to-human communication applications change?
- Will user interfaces become increasingly malleable, personalized, light-weight and asynchronous?
- Will new use cases be enabled when cameras become “point and click” mouse input devices into the real world?
- Will new marketplace products get created because it will become feasible to aggregate demand and/or supply more easily? How will existing marketplaces change?
- What new systems of trust and verification will need to be developed to secure and authenticate these new ‘Human APIs’?
Reach out to: Evan
How will business owners fight back against private equity?
We see two contradictory trends in business ownership: small and mid cap PE firms rolling up small entrepreneur-owned businesses (home services, childcare facilities, fitness, insurance brokerages, healthcare services, etc.); and at the same time, better tools being built to empower solo entrepreneurs.
- How can we create liquidity for business owners outside of PE?
- What new tools might further empower solo entrepreneurs?
- What novel business models (like the Management Services Organization model in healthcare) might emerge to balance autonomy and scale?
- Which industries will be especially enabled by AI automation? Which industries will be somewhat AI-proof, and how?
Reach out to: Finn, Dylan, Danh
How will we balance globalization with national interests while still making progress?
We live in a world characterized by tension between globalization and national interests. The world is increasingly connected, with services more accessible than ever before, creating numerous opportunities for businesses to become borderless. But at the same time we're witnessing a renewed focus on national priorities. This dichotomy increasingly also looks like an opportunity, especially as governments become more open to collaborating with startups.
- How can startups capitalize on services reaching parity across borders?
- How do we manage human capital in a borderless world?
- Which government priorities could startups win? Infrastructure modernization, border security, waste management, disaster response, manufacturing?
- Do we need new ways to interface with the government?
What will simulation-driven product development look like?
Simulations help us understand and predict the world in various domains. Game engines and physics models are improving rapidly and could transform how we understand, predict, and interact with the world.
- How close are game engines to achieving physics-realistic simulations suitable for high-quality data generation?
- In what new domains could these advanced simulations be applied (e.g., personalized medicine, urban planning, climate modeling, etc.)?
- How will simulations impact product development cycles?
- Will AI's reasoning capabilities unlock new frontiers in science?
Reach out to: Finn
Who are the second-order AI beneficiaries?
Beyond the obvious winners of hyperscalers and model providers, who wins? Even if the major foundation model companies capture most of the model value, companies enabling that value will thrive.
- What software helps us build and run larger, more complex data centers? How do we manage energy utilization, permitting, workloads, etc.?
- What happens when compute becomes a commodity?
- How might training data drive growth in synthetic data?
- Who wins in auxiliary services?
Reach out to: Finn
What is the next form of e-commerce?
Shopping and commerce experience online has been menu and search driven. In the offline world, shopping can be fun and a social experience for consumers.
- What new form factors can bring serendipity to online shopping?
- What is the future of commerce search and discovery? How will advertising dollars being spent on Amazon, Alibaba and Flipkart be reassigned?
- How will personal preferences on convenience and functionality create interactive shopping experiences on commerce platforms?
- How will B2B commerce platforms evolve across the various phases of discovery, purchase, fulfillment and service?
How will we balance regulation and customer protection with innovation?
Nations and regulators are struggling to contain risks across finance, information flow, agriculture/food, and healthcare. Governments and regulators are increasingly unsure of which risks they want (or need) to mitigate. The world is getting weirder and the cost of compliance seems to be going up across industries.
- How will businesses stay compliant while staying relevant to customers?
- How can small businesses contain regulatory risk while staying nimble?
- How can large businesses avoid immobility and team bloat?
- How do businesses and regulators interact as risk and compliance requirements evolve more rapidly?
Reach out to: Prateek
What is the next act for Fintech in India?
Fintech was a major winner in the last decade of Indian startup successes, propelled by mobile phones connecting hundreds of millions of people to the internet for the first time. Financial services seem set to explode again as the per capita income increases.
- How will local indic LLMs transform financial services?
- How will service experiences be reimagined? What will be the new emergent models for the newly affluent and mass affluent in the country?
- How will affordability solutions evolve?
- With the UPI transactions crossing 15 trillion monthly transactions and consumers moving online for most financial need states, the number of frauds are going up. A non-digital native population is also vulnerable to agentic phishing. How will risk containment and protection look like for consumers and businesses?
Reach out to: Prateek
How do we address an increase in ennui?
As automation increasingly decouples work and personal identity/meaning, anxiety around financial security and personal identity will also likely grow. Leisure time will increase as agents automate tasks. Post-scarcity will introduce its own problems.
- How do these trends differ between developed and developing economies?
- What new forms of entertainment will emerge out of a radically changed labor and time environment?
- Which services will become more valuable for their meaning-making potential?
Reach out to: Jonathan, Prateek
If you're exploring any of these questions, feel free to reach out to the SPC team member listed—or, better yet, apply to SPC for membership.