The 'laundry war': Sunday Memo claims 99.1% folding autonomy, Weave and 1X align on pricing and subscriptions — laundry becomes THE benchmark for home robots

The "laundry war": Sunday Memo claims 99.1% autonomy in folding, Weave and 1X line up prices and subscriptions — laundry becomes THE benchmark for home robots

🔎 Laundry, the ultimate arbiter of home robots

For ten years, home robots have been sold through edited videos: a shot of the perfect t-shirt, a cut to black on failure. In May 2026, that regime is coming to an end. One player publishes a complete dataset with its error margins, two others display real prices — and a common benchmark emerges to settle the score between them.

The battleground for this fight: your laundry basket. Sunday Memo, driven by its ACT-2 policy, claims 99.1% ±0.3% success across 785 autonomous folds. Weave launches Isaac 1 at $7,999 upfront or $449/month. 1X opens orders for NEO at $20,000 or $499/month. Three strategies, one and the same standardized test.

Why now? Because deformable objects are the last great bottleneck in robotic manipulation — and for the first time, we have a published protocol, public prices, and an evaluation framework to discuss it in something other than marketing terms. All that remains is knowing how to read the numbers. That is exactly the purpose of this article, informed by the pivotal investigation from Biped.News.

The Essentials

• Sunday Memo (ACT-2 policy): 99.1% ±0.3% success (company-reported figures) across 785 autonomous attempts in 9 clothing categories; 778 folds completed, average quality 4.72/5, 98.3% of folds rated 4–5 stars, median of 2 min 13 s per garment.
• Weave Isaac 1: $7,999 cash or $449/month (May 2026, verify on weaverobotics.com) — presented as the first consumer-grade domestic mobile humanoid at this price point.
• 1X NEO: $20,000 ownership or $499/month (May 2026, verify on 1x.tech), 3-year warranty, premium support, Expert Mode with bounded remote human supervision (face blurring, restricted zones, owner approval).
• Success rate is no longer enough: the "basket-to-drawer" benchmark proposed by Biped.News evaluates interventions, intervention minutes, speed, quality, privacy, and cost per useful hour.
• The industry's blind spot: consumer-scale intervention metrics are disclosed by no one. This is THE question to ask before any purchase.

Recommended Tools

Why laundry has become the benchmark that really matters

Because laundry combines everything robotics still struggles with: deformable objects, infinite variability, and long horizons. A rigid object is grasped once; a crumpled t-shirt has to be rearranged with every single movement.

It's no coincidence that the industry chose this test. Rigid object manipulation has long been solved: picking up a cup, placing a box. Deformable objects resist because they partly escape simulation, change state with every contact, and make planning uncertain. Folding concentrates it all: identifying the category, flattening, aligning, chaining folds together, recovering when the fabric slips.

Sunday understood this, and its protocol proves it: 785 autonomous attempts, including 514 starting from a pile placed on a bed, 73 from a basket on a bed, and 198 from a basket on the floor. Three starting conditions, real-world mess — not a demo stage. The result: blouses are the weakest category at 94.7%, while several categories reach 100%.

And there's the showcase effect: everyone has a laundry basket. No need for a fully equipped kitchen or a smart home. It's the universal chore, and therefore the universal ground for comparison.

My opinion is firm: choreographed demos are dead, and that's excellent news. Publishing a protocol becomes the marketing weapon of 2026 — and laundry is the ideal ground to establish this new rule of the game.

Sunday Memo: 99.1% success rate, but read the footnotes

The figure is solid, with a level of transparency rarely seen in this industry — but it's published by the company itself, and it says nothing about human interventions.

The data speaks for itself: 785 autonomous attempts, 778 folds completed (99.1% ±0.3%), an average quality score of 4.72/5, 98.3% of folds rated 4 or 5 stars, and a median of 2 min 13 s per garment. NewMarketPitch cross-checks these figures (778/785, 9 categories) and sums up the situation bluntly: "Sunday has published the industry's only large-scale transparent reliability test."

But the report also owns up to the limitation: consumer-scale intervention metrics are not disclosed. Yet a "successful" fold may have required an intervention before, during, or after — and that's precisely what the general public wants to know before buying.

On the business side, Sunday has raised at least $165M and is preparing a family beta for late 2026, with no commercial price announced. The strategy is legible: win the credibility war first, the price war second.

Publishing its weak points — blouses at 94.7% — is a masterstroke. It forces every competitor to choose: release their own data, or stay in promotional-video mode. The entire industry has just been failed unless it plays along.

Weave Isaac 1 vs 1X NEO: the price war enters the living room

Two business models are going head to head: Weave breaks the price barrier with a ~$8,000 robot, while 1X sells the full-service package at $20,000 with warranty and support. The subscriptions, meanwhile, are almost on par — and that's probably no accident.

According to The Robot Report, Isaac 1 is the first mass-market mobile home humanoid to hit the market at this price. CEO Kaan Dogrusoz stands by it in Business Insider: a ~$8,000 robot aimed first and foremost at laundry folding, significantly cheaper than 1X's NEO (~$20,000).

Spec-wise, NEO comes in at 29.94 kg, with a 68 kg lifting capacity and a 22 dB noise level — figures that remind us we're talking about machines, not gadgets.

Do the math over 36 months: the Weave subscription costs $16,164, twice the $7,999 upfront price. At 1X, the subscription costs $17,964, slightly less than the $20,000 ownership price. In other words: with Weave, buying outright is the rational choice; with 1X, the subscription is the reasonable default if you're not sure you'll stick it out for three years. One important nuance: subscriptions generally include maintenance and support — compare what each plan actually covers before making your call.

The AI market knows this mechanism by heart. When DeepSeek locks in a permanent 75% price cut on V4-Pro and forces everyone else to match it, or when OpenAI opens the GPT-5.6 Sol preview in the middle of a price war, you see the same choreography: price becomes a weapon, the subscription a soft trap, the upfront purchase a commitment. Home robots are copying the LLM playbook.

My prediction: the $50/month gap between $449 and $499 is marketing, not economics. The real fault line is upfront vs. subscription — and above all, what each plan actually includes in terms of service and data.

The "basket-to-drawer" benchmark: the only evaluation grid that matters

Forget the isolated success rate. Biped.News proposes evaluating a home robot on six metrics: interventions, intervention minutes, speed, quality, privacy, and cost per useful hour.

• Interventions: how many times you have to rescue the robot per session. This is the metric nobody publishes.
• Intervention minutes: the actual time you spend on it, stopwatch in hand.
• Speed: minutes per garment — Sunday reports a median of 2 min 13 s, which is a good instinct.
• Quality: the rating of the folds, not just their completeness (4.72/5 at Sunday).
• Privacy: what gets filmed, who watches, where the data goes, and for how long.
• Cost per useful hour: the real price of the hour actually saved.

Let's take a concrete example. A robot at $449/month that saves you 10 hours of unsupervised folding costs about $45 per useful hour. But if it demands five ten-minute interventions per week, you're back in the loop for more than three hours a month — and your real cost climbs accordingly. The formula fits on one line: (subscription + value of your interventions) ÷ hours actually gained.

As long as this metric is missing — and it's missing from everyone, Sunday included — the 99.1% remains a lab figure. A success rate without associated interventions is advertised mileage without fuel consumption.

Take your own measurements (30 days, stopwatch in hand)

The method boils down to three steps. For 30 days, note at each session: task, number of interventions, minutes spent, result obtained. A spreadsheet is more than enough. If you prefer a small online dashboard — an intervention form and a viewable history — hosting at a few euros per month like Hostinger will do just fine.

At the end of the month, calculate your cost per useful hour and compare it with a housekeeper or with your own hourly rate. The result is often brutal. That's precisely why it's useful.

Privacy: the real price of a $499 subscription

NEO's business model relies in part on your data. Learning happens through human teleoperation, and the CEO publicly owns up to it: "without your data, we cannot improve the product" (Bernt Børnich, via Engadget).

The oversight framework is real. In Expert Mode, a human operator supervises remotely, only during scheduled hours; they don't see faces (people are blurred); no-go zones are configurable; no takeover happens without the owner's approval. The robot starts with basic autonomy and improves from there.

This data collection is being industrialized. The Memory Gloves — roughly $200 to produce, with thousands of pairs already shipped according to NewMarketPitch — show that teleoperation hardware is becoming a mass-market consumable. Your living room is turning into a training data production site, with a consent checkbox to tick.

Let me be clear: the blurring and no-go zones are substantial improvements, not cosmetic ones. But the implicit contract remains massive — your gestures, your home, your schedules feed a commercial product. Before subscribing, demand written answers: data retention period, who accesses the feeds, how to limit collection, how to leave with your data.

Weave and Sunday haven't yet published anything equivalent on these issues. That silence is information too.

The groundswell: talent, in-house models, and the race to the living room

This laundry war is no isolated episode. It's the visible tip of an industry that is structuring itself at breakneck pace — talent, proprietary models, and research included.

First signal: the war for talent is entering a brutal phase. Nobel laureate John Jumper is leaving Google DeepMind for Anthropic, while Transformer architect Noam Shazeer joins OpenAI — and the profiles capable of training manipulation policies like ACT-2 are the next front. Robots need the same brains as LLMs.

Second signal: the vertical integration of models. Cloudflare unveils Clef and Clef Flash, its first in-house models; 1X is betting on Redwood AI; Sunday on ACT-2. No one — neither cloud provider nor robot maker — wants to depend on a third party for the core of their product. The embedded model is the new defensive moat.

Third signal: the lab-to-living-room transfer is accelerating. At Stanford, the Homebody project is piloting a humanoid with GPT-6 Astra and a learned zero policy. What gets demonstrated in the lab this year goes on sale in beta the following year — Sunday's family beta is the commercial offshoot of exactly that.

And meanwhile, the generalist models keep climbing: GPT-5.5 dominates the agent leaderboards (98.2), ahead of Gemini 3 Pro Deep Think (95.4) and Claude Opus 4.7 (94.3). My take: household robots are AI's next volume market, and laundry is the gateway in — because it's the universal chore everyone would pay to never do again.

❌ Common Mistakes

Mistake 1: Mistaking 99.1% for full autonomy

What goes wrong: the rate refers to successful folds, not intervention-free sessions. No vendor discloses consumer-scale intervention metrics. The solution: demand "interventions per hour" before any purchase, then measure at home for 30 days.

Mistake 2: Comparing $7,999 and $20,000 without reading the terms

What goes wrong: the $20,000 NEO includes a 3-year warranty and premium support; the Weave subscription comes to $16,164 over 36 months, versus $7,999 cash. The solution: calculate the total cost over your actual horizon (24, 36, 48 months) and list what each plan covers — maintenance, assistance, updates.

Mistake 3: Ignoring the data clause

What goes wrong: Expert Mode, teleoperation, "without your data, we cannot improve the product" — collection is the engine of the product. The solution: set up no-go zones and time windows, verify the blurring, and read the retention policy before your first session.

Mistake 4: Testing on easy laundry

What goes wrong: demos and trials favor t-shirts. Yet blouses top out at 94.7% — and your real laundry basket contains the worst case, not the best. The solution: test on your shirts and your most wrinkled garments before signing anything.

❓ Frequently Asked Questions

How much does Sunday Memo cost?

No commercial price has been announced (May 2026). Sunday has raised at least $165M and is preparing a family beta for late 2026. The reliability figures are public, the price is not — a strategy of winning the credibility war before the price war.

Is Sunday's 99.1% reliable?

It's a company-reported figure, ±0.3% margin, across 785 attempts in 9 categories. NewMarketPitch cross-checks and confirms 778/785 successful folds and an average quality of 4.72/5. But no complete independent replication exists, and interventions are not quantified. Reliable, not definitive.

Isaac 1 or NEO: which one to choose?

If budget comes first, Isaac 1 ($7,999 upfront or $449/month) is half the purchase price. NEO ($20,000 or $499/month) bets on the 3-year warranty, premium support, and learning through teleoperation. Your priority tasks and your tolerance for data collection will settle it.

Does NEO's teleoperation threaten my privacy?

The framework limits the risk: faces blurred for the operator, forbidden zones, scheduled hours, no takeover without your approval. But your data feeds product improvement, as the CEO openly acknowledges. Read the data policy and configure the restrictions before your first session.

When will these robots arrive outside the United States?

1X announces US deliveries in 2026 and other markets in 2027. Weave, a San Francisco startup, has not shared an international timeline. Sunday remains in family beta in late 2026, with no price. For Europe, aim for 2027 and keep an eye on local compliance.

✅ Conclusion

The laundry war already has one merit: it replaces staged demos with dated, debatable figures — it's up to us to demand the one metric that's missing: the cost per useful hour. To dig into the full protocol and the basket-to-drawer benchmark, read the investigation by Biped.News, then time your own interventions: it's the only test that ends in your living room.