Key takeaways
- Investor wishlists reveal funding interests; customer conversations establish demand.
- The operational work around AI creates opportunities beyond the model itself.
- Formal RFS lists, investment theses and market research convey different signals.
- Distribution, workflow ownership and dependable execution deserve early attention.
- A strong founding insight should go deeper than a category investors already agree on.
I love tracking requests for startups as a barometer of what investors want to back and where capital might flow in the following months. It’s not an exact science, of course. VCs are susceptible to herd mentality, and when an influential investor puts a category on the map, more pitch decks and investment committees tend to follow. These lists offer a useful window into investor thinking, even if they tell us much less about what customers will actually buy.
You can track previous editions from 2024, 2025 part 1 and part 2, February 2026, and my most recent Summer 2026 edition.
This edition starts with YC’s Winter 2027 requests, then widens the lens to Pear, a16z, Menlo and QED. Every investor link goes to the original publisher. The investor publications included here date from May 2026 onwards. Research reports and investment theses are labelled accordingly, rather than presented as new accelerator wishlists.
My reading across these publications: the opportunity is increasingly in the difficult work between an impressive AI demonstration and a product that people can depend on. That leaves plenty of room for founders with specific customer insight, even as the headline categories become crowded.
1. Y Combinator: 11 requests for Winter 2027
YC’s latest batch focuses on deployment beyond the model. The following titles and summaries are paraphrased.
- Businesses for cheaper space access. Build for an orbital economy where dramatically lower launch costs unlock previously impractical businesses.
- AI-supported specialist workers. Help generalists perform complex administrative and advisory work previously reserved for experts.
- A domestic robotics industry. Manufacture robots and critical components in America, reducing dependence on imported supply chains.
- Language models with better prose. Develop models prioritising engaging writing, personality and distinctive voice.
- Reusable benefits verification. With consent, connect verified records to multiple public programmes, reducing repeated paperwork and fraud.
- Financing the physical AI economy. Create lending, leasing and risk-management products for robots, compute and infrastructure.
- Models built for specific fields. Deliver specialist performance with substantially lower inference costs and strong domain data.
- AI-powered municipal workflows. Accelerate permits, inspections, licensing and other repetitive local-government processes.
- Unblocking data centre construction. Address shortages across electrical equipment, cooling and essential materials.
- AI inside secure facilities. Enable capable AI deployments entirely within customer premises, including disconnected environments.
- Business-owned models. Help companies train, evaluate and improve models that capture their operational expertise.
Source: Y Combinator, Requests for Startups, Winter 2027; checked 3 November 2026.
2. Pear VC: six requests from founders and operators
Pear’s August 2026 publication asks experienced operators what they wish somebody would build. Its six requests span internal enterprise knowledge and consumer experiences.
- Capture the knowledge that never gets documented. Ali Ghodsi proposes agents that interview employees and turn practical know-how into permissioned organisational context.
- Coordinate AI across the business. Praveen Neppalli Naga wants infrastructure to deploy, govern and improve workflows across departments.
- A coach that learns your patterns. Mario Schlosser describes personalised support that observes behaviour and intervenes when it can help someone change.
- Safe environments for enterprise experimentation. Anahita Tafvizi proposes automatically provisioned testing environments with synthetic data and continuous oversight.
- Help children build with AI. Tomer London wants age-appropriate creation tools that teach underlying concepts, with parental and educational safeguards.
- Evaluate AI against actual business tasks. Andy Fang calls for enterprise-controlled benchmarks and evaluation infrastructure; he also highlights American open-weight models.
For me, the children’s creation platform is particularly interesting. The product challenge is to make the child more capable with each project. Finishing an impressive game is only part of the value; understanding how it works is the other part.
3. a16z speedrun: new interfaces, distribution and ways of working
The 19 May 2026 speedrun newsletter combines four requests from other a16z teams with eight investor ideas. Two, procurement networks and solo-founder tools, appeared in my summer edition. The other ten are worth adding here:
- Earlier melanoma detection: Make timely screening more accessible.
- Orbital defence: Build a prime contractor for protecting space assets.
- Understanding video: Search and extract useful information from recordings and live streams.
- More healthcare applications: Expand clinical capacity and reduce administrative burdens.
- Agents that anticipate needs: Observe context and initiate useful work.
- Accessible agent interfaces: Bring agent capabilities to people who avoid terminals.
- Better judgement tools: Track predictions and improve decision-making.
- Software discovery at scale: Help users navigate an abundance of new applications.
- Products designed for agents: Serve machine users directly.
- Interfaces for participating software: Rethink interaction when software takes action alongside people.
Source: a16z speedrun, That’s a wrap on SR007 applications, 19 May 2026.
4. Menlo Ventures: consumer AI needs to earn greater trust
Menlo’s 2026 State of Consumer AI is research, not a formal RFS. Its forward-looking section identifies five directions: agent-assisted shopping, more habitual agent use, AI as an advertising channel, voice interfaces that complete actions, and practical consumer robotics.
The report’s closing invitation to founders puts trust at the centre. My interpretation is that there is an opening for products that earn progressively greater responsibility: first helping users evaluate a decision, then carrying it out with clear controls. In commerce, that makes consent, payment security and dispute resolution product features with direct commercial importance.
Source: Menlo Ventures, 2026: The State of Consumer AI.
5. QED: AI that connects the financial life
QED’s May 2026 wealth-management thesis highlights fragmented advisor software, growing financial complexity and the potential to expand advisory capacity. It sees opportunities in unified financial data, automated administration, more accessible personalised advice and infrastructure for broader asset coverage.
This is a sector thesis rather than a numbered RFS. My reading is that the product opportunity goes beyond answering financial questions: useful systems need to connect accounts, prepare workflows and help advisors maintain an accurate picture of a household. Reliability and the customer relationship remain central.
Source: QED Investors, AI is rebuilding wealth management, 22 May 2026.
What I would take from these lists as a founder
There is a temptation to read an RFS as an instruction to enter a category. I would use it as a starting point for customer research. An investor naming a problem tells you there may be interest in funding a solution. It does not establish that the buyer has a budget, that procurement is manageable, or that your team has an advantage.
Three questions help separate an appealing theme from a company worth building. Who pays for this problem today? What can you do materially better than their current alternative? What improves as your company gains customers that a competitor cannot simply reproduce by using the same model?
For Israeli founders, I would look for a concrete connection between prior experience and a customer’s unresolved problem. Secure deployment, enterprise workflows, creative tools and agentic commerce are all worth exploring, but the initial opportunity needs to be much more specific than the category name. The first customer should be able to recognise their own problem in the pitch.
At Remagine Ventures, we remain open for business and actively back Israeli founders at pre-seed. I’m particularly interested in how AI changes what people create, how they buy and how small teams operate. These lists are a useful conversation starter. The most compelling founders will bring their own insight into what the lists have missed.
He is a former General Partner at Google Ventures (GV) in Europe, former head of Google for Entrepreneurs in Europe, and founding head of Campus London, Google's first startup hub. Eze writes on Israeli tech, venture capital, artificial intelligence, and founder strategy.
He is also the founder of Techbikers, a nonprofit that brings together the startup ecosystem on cycling challenges in support of Room to Read.
- Requests for Startups: Winter 2027 edition - November 4, 2026
- Big ideas, basic questions - November 3, 2026
- Weekly Firgun Newsletter – October 30 2026 - October 30, 2026
