a16z Speedrun's SR007 Just Opened: The 5-Week-ARR Bar and 6 Plays Indie Hackers Can Ship
a16z Speedrun opened SR007 applications through May 17, 2026. Inside the new ARR bar Bota and Bilrost just set, the patterns hidden in their 2026 portfolio, and the 6 plays indie hackers can ship without applying.
Key Takeaways
- a16z Speedrun's SR007 applications are open April 20 – May 17, 2026, with the 12-week program running July 27 – October 11 and Demo Day on October 6. Speedrun has now deployed $300M+ across 250+ companies at a 0.4% acceptance rate.
- The signal is no longer the public '14 Big Ideas' list — it's the portfolio. Bota hit $710K ARR in five weeks. Bilrost hit $680K ARR in five weeks. Cascade did $201K ARR in three months. Five-figure-MRR-by-Demo-Day is the new venture-fundable bar.
- Six patterns dominate the 2026 portfolio: agent-native infrastructure, AI-native vertical service businesses, prompt-free consumer apps, voice agents going full-lifecycle, AI startups selling to AI startups, and 'see me' personal AI.
- The six indie hacker plays mirror those patterns and are all shippable from a laptop in 30–60 days. You do not need to apply to act on the thesis — but if you do, the signal SR007 wants is a working product with paying customers, not a deck.
a16z Speedrun's SR007 batch is now open — applications close May 17, 2026, the program runs July 27 – October 11, and Demo Day is October 6. With SR007, Speedrun has now deployed more than $300M across 250+ companies at an acceptance rate under 0.4%. The public “14 Big Ideas” list everyone reads tells you what they think. The portfolio tells you what they fund. The two are not the same — and the gap is where indie hackers should be paying attention.
The Real Signal: A New ARR Bar Hidden in the Portfolio
Read Speedrun's SR006 portfolio descriptions and one number keeps appearing: weeks, not quarters, to first revenue. Bota, an agent-to-real-world systems platform, hit $710K ARR in five weeks. Bilrost, automating commercial loan processing, did $680K in the same window. Cascade, bringing AI to construction, cleared $201K ARR in three months. These are not cherry-picked — they are the public details Speedrun chose to highlight.
That is the new bar. Two years ago, an accelerator-fundable team showed up with a deck and a working prototype. In 2026, the bar has moved to a working prototype with paying customers and traceable ARR by the time you apply. Speedrun's 0.4% acceptance rate is not a brutal filter on ideas — it is a filter on velocity. Founders who can wire AI into a real workflow, get it in front of buyers, and bill them in five weeks are the ones getting funded.
For indie hackers the read is liberating, not discouraging. The same skills that produce a Speedrun-fundable company also produce a bootstrapped business that does not need Speedrun. Five-week ARR is just “ship the wedge, charge for it, stay alive.” That is the indie hacker playbook — newly validated by the most signal-rich consumer/AI accelerator on the market.
What Speedrun Has Actually Been Funding in 2026
Ten companies from the most recent cohort, sorted by what they reveal about the thesis. Four buckets emerge cleanly: agent infrastructure, AI-native vertical service businesses, AI orchestration for the enterprise, and “see-me” consumer products. Notice how few are pure consumer apps and how many are AI replacements for analog service work.
| Company | What it does | Bucket | Metric |
|---|---|---|---|
| Bota | Connects AI agents to real-world systems | Agent infra | $710K ARR in 5 weeks |
| Bilrost | Automates commercial loan processing | AI-native vertical | $680K ARR in 5 weeks |
| Cascade | AI for construction operations | AI-native vertical | $201K ARR in 3 months |
| Advocate | AI-native care coordination | AI-native vertical | $7M projected revenue |
| Acceler8 | Workforce intelligence platform | AI orchestration | $700K ARR |
| Cedar | Agents challenging LinkedIn (ex-HootSuite) | AI-native vertical | Pre-launch |
| Auto | AI camera turning photos into personal apps | See-me consumer | Consumer launch |
| Alike | Agent collaboration layer for enterprises | Agent infra | Pre-launch |
| Amdahl | AI context layer for enterprise GTM (ex-Databricks) | AI-native vertical | Pre-launch |
| August | AI autonomous banking | AI-native vertical | Pre-launch |
Pulled from a16z Speedrun's public portfolio page. Buckets are our read.
The 6 Plays Indie Hackers Can Ship Without Applying
Each play maps to a pattern in the 2026 portfolio plus the freshest threads from a16z's own Big Ideas Parts 2 and 3. None requires a SAFE, a co-founder, or a visa. All are buildable in 30–60 days with a coding agent and a customer list.
1. Agent-Native Tools for One System (Bota pattern)
Bota's wedge is connecting agents to real-world systems. The category is enormous because every business runs on systems agents cannot use well — CRMs, ERPs, payroll, ticketing, inventory. Pick one, ship the agent-native interface, charge per call.
What to actually build
- Pick one painful enterprise system — NetSuite, Workday, ServiceNow — and ship a clean MCP server plus per-agent auth. Charge by invocation, not by seat.
- Sell to teams already deploying agents — they have budget and they hit the “agent does not know our system” wall every day.
- Distribution wedge: become the default in Claude, Cursor, or ChatGPT's app marketplace. Whoever wins “default ServiceNow MCP” gets agent traffic for free.
2. AI-Native Vertical Service in a Boring Industry (Bilrost / Cascade pattern)
Three of Speedrun's fastest-growing recent bets — Bilrost in commercial loans, Cascade in construction, Advocate in care coordination — are the same shape: a single regulated, paperwork-heavy workflow that AI can now run end to end while a human signs off. Software margins, service-business stickiness.
What to actually build
- Pick one regulated workflow in an industry you know — sales tax filings, R&D credits, lien releases, prior-auth in healthcare. Own that single slice for a single buyer profile.
- Quote a flat fee 40–60% under the incumbent firm. Use AI for the drudge, keep one human in the loop for sign-off. The compounding moat is your in-house playbook — by customer 50 the AI does 90% of the work.
- Sell, do not market. Cascade's $201K ARR did not come from content. It came from cold outreach to construction GMs with a working demo.
3. Prompt-Free Consumer Apps (Auto pattern + “See Me” thesis)
Marc Andrusko called it the death of the prompt box. Bryan Kim called it the shift from “help me” to “see me.” Auto, the ex-Snapchat team in SR006, is the proof of concept: an AI camera that watches what you do and generates a personal app for it. The category is wide open below them.
What to actually build
- Pick one repetitive consumer behavior — meal logging, receipts, gym sets, kid moments — and ship a camera-first app that auto-acts. No chat. No prompts. Tap to capture, AI does the rest.
- Distribution: ChatGPT's Apps SDK with its 900M users. Anish Acharya wrote that it is the new App Store. He is right — and the review queue is shorter than Apple's.
- Pricing wedge: $5/month for the personal version, $20/month for family/team. Beat the freemium-with-ads model the consumer-AI incumbents are stuck on.
4. Voice-First Vertical Agents (Olivia Moore thesis)
Moore wrote in Big Ideas 2026 Part 2 that voice agents are graduating from scheduling to full customer-lifecycle management. The pattern in Speedrun's portfolio confirms it: the early voice plays were horizontal (book my meeting), the funded plays are vertical (run my dental front desk).
What to actually build
- Pick one inbound-call-heavy vertical — independent dental, auto repair, real estate brokerages, legal intake, HVAC. Build the agent that handles the first 80% of calls and hands off the rest to a human.
- Charge per booked appointment, not per minute. The buyer knows the unit economics of a customer; meet them there.
- The moat is the playbook — what to say when a patient is angry, when a quote is too high, when a lead is hostile. That data compounds with every call and is impossible for a horizontal voice tool to replicate.
5. Sell to AI-Native Startups (James da Costa thesis)
Big Ideas 2026 Part 2 has a quiet thesis from James da Costa: AI startups serving other AI startups are reaching scale because their customers are the fastest-growing companies in the market. Cedar (selling to AI-native HR teams), Amdahl (selling to AI-native go-to-market teams), and Alike (selling to companies rolling out agent fleets) all sit in this lane.
What to actually build
- Pick one AI-native function — agent observability, prompt version control, eval-as-a-service, agent identity (Sean Neville's “Know Your Agent”), token-spend optimization.
- Your customer list is public. a16z Speedrun, YC, and South Park Commons portfolios. Build for them — they ship in days, not quarters, and they pay invoices fast.
- Pricing: usage-based, ramping with their growth. When your customer becomes a unicorn, you become a real company.
6. Substantive Research-as-a-Service (Kominers thesis)
Scott Duke Kominers's entry in Big Ideas 2026 Part 3 is the most slept-on: AI can now do doctoral-quality research on abstract instructions. The opportunity is not a horizontal “Deep Research clone.” It is vertical: produce better research than the McKinsey associate, the patent attorney, or the equity analyst, in one industry, faster, for a fraction of the price.
What to actually build
- Pick one research-heavy task a senior professional does manually — patent prior-art searches, M&A target memos, clinical literature reviews, competitor teardowns, due-diligence briefs.
- Charge per delivered report, not per hour. The buyer's reference price is “what would a $400/hr consultant charge,” and you should be 70% under that — still huge margins for you.
- The moat is the corpus. Your private database of every report you have produced — annotated, corrected, validated by the customer — is the asset. Generic Deep Research cannot match it.
Stay Ahead of the Trends
Get founder-grade breakdowns of every YC RFS, accelerator drop, and market shift — delivered weekly. No fluff.
What Speedrun Is Quietly Skipping (and Why That Matters)
The portfolio is as informative for what is missing as for what is present. Three categories that dominated 2024–2025 startup discourse have nearly vanished from the recent cohorts: pure ChatGPT wrappers, generic horizontal agent frameworks, and AI productivity tools without a vertical. The implication is plain — these are no longer fundable shapes, and they are no longer defensible to bootstrap either.
The funded shapes share three traits: a vertical with real workflow stakes (loans, construction, care, dental), a customer who can be cold-emailed in days (not Fortune 500 procurement), and a unit of value priced for outcomes (booked appointments, processed loans, completed reports). If your idea fails any of those three, the portfolio data is telling you to keep iterating.
Apply to SR007 or Don't — The Plays Work Either Way
Applications close May 17, 2026. If you have a working product, paying customers, and traceable ARR, the Speedrun terms are unusually founder-friendly: $500K for 10% upfront via SAFE, another $500K committed for the next round, more than $5M in cloud and AI credits, no board seat, and a 12-week in-person sprint in San Francisco. The acceptance bar is what they look at in your application, and the Speedrun team has been explicit on three things: velocity (how fast you ship and learn), earned insight (a non-obvious thesis you can defend), and customer conversations already underway by the time you apply.
If you are not applying, none of the six plays above need a SAFE to ship. The portfolio is the thesis; the public “Big Ideas” list is the press release. Build the wedge in 30 days, charge for it in week three, and you are running the same playbook as the SR006 cohort — just without the $5M in credits and the equity dilution.
Tools to Validate Your SR007-Inspired Idea
Before you build, score the idea, sharpen the wedge, and turn it into a prompt your coding agent can ship from.
Related reading: a16z Speedrun's 14 Big Ideas for 2026 — our deep-dive on the December 2025 substack post, with each idea translated into a concrete indie hacker play. Useful as a companion to this piece.
The Bottom Line
- SR007 applications close May 17, 2026. Speedrun has now deployed $300M+ across 250+ companies. The acceptance rate sits below 0.4% and the bar is no longer a deck — it is a working product with paying customers.
- The portfolio is the real thesis. Bota and Bilrost both hit ~$700K ARR in five weeks. AI-native vertical service businesses, agent infrastructure, and prompt-free consumer apps are the funded shapes.
- Six plays are reachable from a laptop. Agent-native MCP servers, AI-native vertical service in a regulated workflow, prompt-free consumer apps, voice agents going full lifecycle, sell-to-AI-native, and substantive research-as-a-service.
- Apply or don't — the playbook is the same. Pick one play, ship the wedge in 30 days, charge in week three, and you are running the Speedrun-grade playbook either way. The accelerator just adds rocket fuel.
Sources
Don't Miss the Next Big Shift
Every week, we break down the trends that matter for indie hackers and SaaS founders — accelerator drops, model launches, market shifts, and the playbooks the founders ahead of the curve are already running. Join the list.
Join 3,000+ founders who stay ahead of the curve