Sunday Robotics is a robotics company based in Mountain View, California, that develops a household robot called Memo together with the foundation models that control it. It was founded in 2024 by Tony Zhao, who left a computer-science PhD programme at Stanford, and Cheng Chi, in what Business Insider describes as a Silicon Valley hacker house; Zhao is chief executive and Chi chief technology officer. The company emerged from stealth on November 19, 2025 (Source: finance.yahoo.com; businessinsider.com).
The founders' prior research includes ALOHA, transformer adaptation for robotics, and diffusion policy. Zhao's stated reason for choosing the home over the factory is that the home is messy and unpredictable, and that a robot able to handle it is the clearest path to what he calls physical artificial general intelligence (Source: finance.yahoo.com; businessinsider.com).
Snapshot
Valuation and funding
| Date | Amount | Round / event | Source |
|---|---|---|---|
| 2026 (first half) | $1.15B | $165M raised | (Source: businessinsider.com) |
| 2025-11-19 | Not disclosed | $35M from Benchmark and Conviction, disclosed at emergence from stealth | (Source: finance.yahoo.com) |
Headcount
| Date | Employees | Source |
|---|---|---|
| 2026-07 | more than 100 | (Source: businessinsider.com) |
| 2025-11 | 25 engineers and researchers | (Source: finance.yahoo.com) |
Conviction's Sarah Guo wrote the company's first cheque. Benchmark general partner Eric Vishria, announcing the firm's investment, framed the constraint as data rather than hardware: "The promise of AI robotics isn't back-flipping or dancing demos, but robots that work in messy, real-world situations. To have those, we need real-world training data. We have about one-millionth of the data we need" (Source: finance.yahoo.com; businessinsider.com).
Memo and the Skill Capture Glove
Memo is a wheeled rather than bipedal robot, a choice the company attributes to balance, lower weight, and stability during a power loss; its exterior is silicone-clad. At the November 2025 launch Sunday described Memo as trained on approximately 10 million episodes of household routines collected in more than 500 real homes using a wearable it calls the Skill Capture Glove, which records how people move, clean and organize (Source: finance.yahoo.com).
The glove is the company's data-collection mechanism and its stated cost advantage: it costs about $200 to make and mirrors the shape and sensor layout of Memo's hands, so that recorded human movement translates more directly into robot movement (Source: businessinsider.com). The approach sits between the crowdsourced consumer-app collection Figure describes for its Index dataset and the hand-held UMI collection Skild AI rates as a middle option on hardware proximity (Introducing S1: In-Context Learning for Robotics).
Models
ACT-1
ACT-1 is the earlier model, demonstrated clearing a dinner table, loading a dishwasher, and pulling a shot of espresso (Source: businessinsider.com).
ACT-2
Sunday unveiled ACT-2 on July 16, 2026; the company's own preview post is dated July 17 (Source: businessinsider.com; sunday.ai). The company says the model learns new behaviors from a single demonstration and carries what it learns in one environment into the next, avoiding retraining for each home. Its headline figure is that Memo folded laundry successfully more than 99% of the time when tested in unfamiliar homes and on garments it had not been specifically trained to handle; Sunday says it spent the preceding months testing in employees' homes and in Airbnbs, and Zhao states that "Memo's performance in our office almost perfectly transfers to its performance in the wild" (Source: businessinsider.com; sunday.ai).
Training is described in two stages: the model first learns basic physical skills from people performing everyday tasks while wearing the sensor gloves, then improves from its own mistakes using Sunday's in-house robot fleet, with new skills acquired within minutes from that base (Source: businessinsider.com).
Zhao told Business Insider that "2026 will be when the timeline for autonomous home robots shrinks in a very drastic way because of this breakthrough." The claims are the developer's own and no third-party evaluation had been reported. Sunday states that vacuuming, organizing toys, fastening zippers and making coffee are being learned but are not yet reliable enough (Source: businessinsider.com).
The "Solve" reporting proposal
Alongside ACT-2, Sunday proposed an industry reporting convention it calls a Solve, intended to separate reliable capabilities from demonstrations that omit how many attempts a robot needed or the conditions under which it was tested. Under the proposal a company would state what a robot can do, where it was tested, and how much additional training or human assistance it requires in each new setting. Zhao's stated rationale is that "each demo is a different robot, setting, and object, and it's really hard to figure out if we're progressing." Sunday counts laundry folding as Memo's first Solve capability (Source: businessinsider.com). The proposal addresses the comparability problem visible across the 2026 embodied-model releases, where GEN-1.5 and S1 report on separate internal suites under different measurement conventions (GEN-1.5: Embodied Foundation Models are One-Shot Learners; Introducing S1: In-Context Learning for Robotics).
Deployment and business model
Sunday opened applications for a Founding Family Beta on November 19, 2025, for fifty households receiving individually numbered robots, with the programme stated to launch in late 2026. In July 2026 the company said it planned to place Memo in homes through a beta programme that autumn, with Zhao declining to say how many homes would participate (Source: finance.yahoo.com; businessinsider.com).
Memo is stated to operate autonomously, with a remote operator stepping in only when a customer asks for help, an arrangement Zhao compares to the remote assistance used by Waymo's self-driving cars. Sunday says it does not plan to use those interventions to collect training data from customers' homes: "Being fully autonomous is the best way to build trust. We are not trying to creep in a teleoperator into your home to collect your data so that we can get our model to be better" (Source: businessinsider.com). That position distinguishes Sunday's data strategy from developers whose pipelines depend on deployment-time collection.
Competitive position
Business Insider places Sunday among several companies preparing to put robots in homes: Weave Robotics, which planned to begin shipping the $7,999 wheeled laundry-folding Isaac 1 in California in autumn 2026; 1X, aiming to deliver its $20,000 NEO humanoid around the same time; and Tesla, working toward bringing Optimus into homes (Source: businessinsider.com). The distinguishing claims in Sunday's July 2026 release are the reliability figure on unfamiliar garments in unfamiliar homes and the glove-based collection method, rather than single-demonstration learning itself, which Generalist and Skild AI also claimed in August 2026.
Relationships
- instance-of: AI Robotics — a home-robot developer building its own control models.
- depends-on: In-Context Learning — the single-demonstration claim ACT-2 rests on.
- depends-on: Foundation Models — the pre-train-once, deploy-many pattern applied to household tasks.
- related: Figure AI, Skild AI, Generalist, Tesla — peer developers of embodied models, and the contrasting data-collection strategies.
- related: GEN-1.5, S1 (Skild AI) — concurrent single-demonstration claims measured on separate internal benchmarks, the comparability problem the Solve proposal addresses.
Open questions
- Whether the more-than-99% laundry-folding figure holds outside Sunday's own testing; no independent evaluation has been published, and the company has not published the ACT-2 test protocol in a form that would let one be run.
- Whether any other developer has adopted the Solve reporting convention.
- How many households the autumn 2026 beta placed Memo in, and how the fifty-household Founding Family Beta announced in November 2025 relates to it.
- Memo's price. Sunday has not published one; secondary sources report a build cost near $20,000 and a stated target below $10,000, neither confirmed here from a company statement.