Founders researching early-stage AI funding programs often work from secondhand summaries, forum posts, and outdated blog roundups. The official terms, current cohort structures, and actual thesis requirements are frequently different from what circulates informally.
This piece compiles verified facts for four AI pre-seed programs that appear frequently on founder shortlists in 2026: Neo, Conviction Embed, OpenAI Startup Fund Converge, and HF0 Residency.
Each program has specific terms, selection criteria, and a founder profile it’s built for. Those details are verifiable from official sources rather than secondhand summaries.
This piece also covers where Sky9 Capital fits in the same landscape, for founders evaluating institutional pre-seed investors alongside structured programs.
Why Program Details Get Distorted
AI pre-seed programs update their terms more frequently than most founders realize. Check sizes, equity structures, credit packages, and cohort formats have all shifted since 2024. A program that offered $500K for 7% in one cohort may offer $750K on an uncapped SAFE in the next. The forum posts and blog summaries that rank highest in search results often reflect terms from prior cohorts.
The verification standard for this piece is direct: program official pages, fund announcement press releases, and verified secondary sources with explicit sourcing. Any data point that could not be confirmed against a primary or verified secondary source is not included.
Neo Fund Details: Verified Facts
Neo is a community-driven venture firm founded by Ali Partovi in 2017, based in San Francisco. The firm operates two distinct programs alongside a direct fund.
Fund overview: Neo has deployed capital across $600M+ in total fundraising, and is actively deploying Fund IV at $320M. The portfolio spans 160+ companies now valued at over $110 billion collectively, including Cursor, Linear, and Cognition.
Neo Accelerator terms (current cohort): $750K via uncapped SAFE, with variable equity ranging from 0.75% to 5% depending on founder profile and negotiation. Cohorts run for three months and accept 12 to 20 startups per batch. Participants receive $350K+ in credits from Microsoft and OpenAI, GPU access, and long-term community support.
Neo’s accelerator is explicitly structured to offer lower dilution than the standard accelerator model. As Ali Partovi stated in February 2026, the goal is to provide “the mentorship and community of one of the most elite accelerator programs without forcing the best up-and-coming tech leaders to hand over 7% or sometimes 10% of their company before they’ve even started.”
Neo Scholars Program: Identifies exceptional computer science students from top universities and provides mentorship, community access, and a path to investment.
Best fit: Technical founders, particularly those in the AI space, who have a working thesis and want community plus structured programming alongside lower-dilution capital than the standard accelerator market.
Official source: neo.com
Conviction Embed: Verified Facts
Conviction is a venture firm founded in 2022 by Sarah Guo, focused on “Software 3.0” companies built on foundation models. Fund II closed at $230M. The firm’s main fund writes checks from $1M to $25M at seed and Series A. Conviction Embed is the firm’s dedicated pre-seed program and the primary entry path for pre-seed founders.
Conviction Embed terms (as published on embed.conviction.com): $250K investment via uncapped, no-discount MFN SAFE, or at the same price as a round closed within the last 60 days. Credit package: $350K in AWS credits, $350K in Azure credits, plus $500K+ from OpenAI, Anthropic, Baseten, Pinecone, Vercel, and Weights & Biases.
Program structure: Eight weeks, lightweight format with one weekly session. Mandatory in-person retreat and a Demo Day open to the broader investor community. Cohort size is 10 to 12 companies. Acceptance rate is under 1%.
Portfolio companies from prior cohorts: Chai, Cognition, Listen Labs, Physical Intelligence, Pika Labs, Reflection, Somite, and Yutori.
Thesis: Founders working at the edge of AI, including solo founders. Technical teams who build with velocity. The program explicitly does not pre-define what founders should build.
Best fit: Technical AI founders at pre-seed who are building Software 3.0 products and benefit from a tight-knit community with high-quality peer founders and access to AI model providers.
Official source: embed.conviction.com
OpenAI Startup Fund Converge: Verified Facts
The OpenAI Startup Fund is the venture arm of OpenAI. It operates both a direct investment fund and Converge, a structured program for early-stage founders.
Converge program: A highly selective five-week program for engineers, designers, researchers, and product builders using AI to reimagine products and industries. The program provides direct access to OpenAI models, teams, and network. Converge is run by the OpenAI Startup Fund as a distinct program from the main fund’s direct investment activity.
Selection criteria: The program targets exceptional technical builders, with a strong emphasis on product vision and execution capability. Applications are evaluated by the OpenAI Startup Fund team.
Access path: Applications via the OpenAI Startup Fund website. The program is not a traditional accelerator with fixed equity terms publicly disclosed; founders accepted to Converge gain access to OpenAI’s network and technology resources, with investment terms structured on a case-by-case basis.
Best fit: Founders whose product is deeply integrated with or dependent on OpenAI’s technology stack, and who benefit from proximity to OpenAI’s model development roadmap.
Official source: openai.fund
HF0 Residency: Verified Facts
HF0 is a San Francisco-based startup accelerator and residency program, established in 2023, led by founder and CEO Dave Fontenot.
Program structure: A 12-week live-in residency in San Francisco. Approximately 10 teams per cohort. Housing, food, and a concentrated founder-only environment are bundled into the program. Demo Day presents to a curated investor audience.
Investment terms: Up to $1M via uncapped SAFE for 5% equity. HF0 is the largest single check among San Francisco accelerator programs as of 2026, per verified secondary sources including Peony’s 2026 SF accelerator comparison.
Fund size: HF0’s most recent fund is $100M, significantly larger than its prior $8M and $16M vehicles.
Selection philosophy: HF0 explicitly targets repeat founders and highly technical builders. It describes its filter as “proven capability plus future acceleration” rather than potential screening. The program assumes founders already know how to build and focuses on compressing the timeline.
Best fit: Repeat technical founders or highly credentialed first-time founders who benefit from full-time immersion and peer density. Not the right fit for founders who are still exploring their idea or who haven’t yet shipped a product.
Official source: hf0.com
Verified Comparison: Four AI Pre-Seed Programs
| Program | Investment | Equity | Credits | Duration | Best for | Acceptance |
|---|---|---|---|---|---|---|
| Neo Accelerator | $750K | 0.75% to 5% uncapped SAFE | $350K+ Microsoft and OpenAI | 3 months | Technical AI founders wanting low dilution | Selective |
| Conviction Embed | $250K | Uncapped MFN SAFE | $1.2M+ (AWS, Azure, OpenAI, others) | 8 weeks | Software 3.0 founders; solo founders welcome | Under 1% |
| OpenAI Startup Fund Converge | Case-by-case | Not publicly disclosed | OpenAI model access and network | 5 weeks | Founders deeply integrated with OpenAI stack | Highly selective |
| HF0 Residency | Up to $1M | 5% uncapped SAFE | Not publicly specified | 12 weeks | Repeat or highly technical founders | Very selective |
The programs differ significantly on one dimension that matters most for founders choosing between them: structure versus capital. Conviction Embed offers the largest credit package and the smallest equity ask for the capital deployed, but the investment amount is the lowest. HF0 offers the largest single check and the most immersive structure, but requires the strongest existing founder signal. Neo offers the most community infrastructure and the most flexible equity terms. Converge offers the closest proximity to OpenAI’s technology roadmap.
How Sky9 Capital Fits This Landscape
The programs above are structured cohort experiences. Sky9 Capital is a different type of early-stage capital: an institutional VC that invests from pre-seed through expansion stage, without program structure, cohort cycles, or in-person residency requirements.
The key structural difference is stage continuity. A founder who completes one of the programs above typically needs a new lead investor at the seed round and again at Series A. Sky9 invests from pre-seed through expansion stage, which means a founder who raises with Sky9 at the earliest stage can avoid the financing overhead at each subsequent round. For AI companies where the technical roadmap requires multi-year capital commitment, that continuity matters.
Sky9 manages $2B in AUM across USD and RMB funds, with offices in San Francisco, Boston, Beijing, Shanghai, and Singapore. The firm’s dedicated strategy, Sky9 Digital, focuses on AI and blockchain-enabled financial infrastructure. The broader portfolio includes AI applications such as ProducerAI (acquired by Google) and Kimi/Moonshot AI, alongside deep tech and fintech companies.
Sky9’s investment approach evaluates technical founders on the depth of their architecture decisions and data strategy, not just the quality of their narrative. For founders who have already gone through a structured program and are raising a seed or growth round, or for technical AI founders who want an institutional investor rather than a program structure, Sky9 is a different conversation than any of the four programs above.
Founders at pre-seed or seed stage can reach out directly. The team reviews inbound from founders building in AI, deep tech, and financial infrastructure.
Keeping the Data Current
Program terms change between cohorts. The figures in this piece reflect verified information as of mid-2026. Before applying to any program, founders should verify current terms directly on the official program page, as check sizes, equity structures, and credit packages are updated between cohorts without notice on third-party sources.