Industries / Consumer goods / Consumer appliances
Design
Deciding what a home appliance does and how people use it: the product itself, its controls, its connections to phones and networks, and the safety of its design.
5
Milestones achieved, from 1882 to 2023.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1882
An iron heated by wire
On 6 June 1882 the US Patent Office granted Henry W. Seely of New York patent 259,054 for an electric flat-iron. A resistance, a wire that heats when current runs through it, sits sealed inside the iron and heats its face, and a layer of material that does not conduct heat keeps the warmth at the face.
Credit
Henry W. Seely
1886
A dishwasher that sprays
On 28 December 1886 Josephine Cochrane (also spelled Cochran) of Shelbyville, Illinois, was granted US patent 355,139 for a dish-washing machine. Dishes sat in wire compartments in a wheel inside a copper boiler, and a motor turned the wheel while hot soapy water was pumped over them. The National Inventors Hall of Fame calls it the first machine with racks made to hold dishes in place and the first to clean with water pressure instead of scrubbers. It won a prize at the 1893 World's Columbian Exposition in Chicago, and her company later became part of KitchenAid, which brought out a dishwasher based on her design in 1949.
LessonThe first buyers were hotels, large restaurants, hospitals and colleges; the Hall of Fame says households did not buy it at first.
Credit
Josephine Cochrane
1967
Microwaves reach the countertop
Percy Spencer of Raytheon filed a patent on 8 October 1945 for cooking food with microwaves from a magnetron, the tube that makes the microwaves; it was granted in 1950. The patent reports that a hard-boiled egg took about 2 kilowatt-seconds of microwave energy, against 36 kilowatt-seconds by ordinary cooking. The first Radarange ovens were large, water-cooled machines for restaurants. IEEE Spectrum calls Tappan's RL-1 of 1955, a wall oven priced at US$1,295, the first consumer microwave oven. In 1967 Amana, an appliance maker Raytheon had bought about 1965, put Radaranges on household countertops for US$495.
LessonA new kind of appliance can reach homes before its safety data does: in March 1973 Consumers Union said none of 15 ovens it tested could be called "completely safe", partly because there was no solid data on safe radiation levels (IEEE Spectrum).
Credit
Percy Spencer and Raytheon; Tappan; Amana
2020
A standard for usable controls
The International Electrotechnical Commission (IEC) published IEC 63008 on 27 March 2020, written by its committee on the performance of household appliances (TC 59). It sets accessibility requirements for knobs, buttons, touch controls, display screens, doors, lids, drawers and handles, so that older people and people with disabilities can use them. It leaves out control by phone apps and networks, speech and gestures, and it states that it does not allow a full assessment of an appliance's overall accessibility.
Credit
IEC Technical Committee 59
2023
Appliances join a shared standard
On 23 October 2023 the Connectivity Standards Alliance released Matter 1.2, the second update to Matter, its smart-home connection standard. It added nine device types, among them refrigerators, dishwashers, laundry washers and robotic vacuums; dryers were left for a later release. The Alliance says the appliance work gives a base of shared features, such as temperature monitoring and status notifications, for appliances to come.
Credit
the Connectivity Standards Alliance's Matter Working Group
Watch for
Gemini looks inside the fridge
Samsung says an update built with Google's Gemini model lets the cameras in its Bespoke AI Family Hub refrigerators recognize more kinds of food, fresh and packaged, "with greater accuracy", on top of the 40 food types they already recognized. The update reached select US models with 32-inch screens in May 2026, after launching in South Korea; models with 9-inch screens follow later. Samsung gives no accuracy figure.
Achieved whenan outside tester stocks several of these fridges with a published list of foods and reports the share the camera names correctly, fresh and packaged items counted separately.
Claimed by
Samsung, May 2026, as reported by SamMobile
Open
01
Blind testers grade appliance controls
A filmed, scored test in which 10 to 20 blind or low-vision people each try the same five everyday tasks on 20 current appliances bought at retail: start a 40 °C cotton wash, set an oven to 180 °C, start a dishwasher's eco program, defrost on a microwave, and switch the appliance off. Each task is tried twice, on the appliance's own panel and through its phone app, and stops before anything heats food. Each control is also scored against the IEC 63008 checklist. Publish the success rate, time and errors for every model.
The briefCredit · unclaimed02
AI reads the control panel
An open set of at least 500 photos and short videos of real appliance control panels (washers, dryers, ovens, hobs, microwaves, dishwashers) in homes and shops, with every control named and the panel's current setting written down. Use it to test current vision-language models, AI models that answer questions about pictures, on two jobs: say what the appliance is set to now, and guide a fingertip to a named control. Then test the best model with blind users.
The briefCredit · unclaimed
Open 01 · the brief
Blind testers grade appliance controls
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A filmed, scored test in which 10 to 20 blind or low-vision people each try the same five everyday tasks on 20 current appliances bought at retail: start a 40 °C cotton wash, set an oven to 180 °C, start a dishwasher's eco program, defrost on a microwave, and switch the appliance off. Each task is tried twice, on the appliance's own panel and through its phone app, and stops before anything heats food. Each control is also scored against the IEC 63008 checklist. Publish the success rate, time and errors for every model.
Why it is still open
The evidence so far comes from surveys and small studies without scores per model. Vision Australia's 2026 brand report card rests on its 2025 survey of blind and low-vision consumers and names brands people found more usable, such as Bosch with its Home Connect app; it publishes no measured task success. An et al. (2024) watched 12 visually impaired people use six appliances in a home-like room and listed design problems, without success rates per model. IEC 63008 covers controls but leaves out apps and speech and does not rate a whole appliance.
Built with what exists
Appliances on a showroom floor or bought and resold; a written protocol drawn from IEC 63008; phone cameras for filming; the iOS VoiceOver and Android TalkBack screen readers for the apps; a shared spreadsheet published under an open license.
First step
Write the five task scripts and the scoring sheet, and try them on two appliances with two testers recruited through a local blindness organization.
Done looks like
20 models, 5 tasks and at least 10 testers, with consented videos, success rates and median task times published per model, and every video scored by a second rater with the agreement between raters reported.
Open 02 · the brief
AI reads the control panel
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open set of at least 500 photos and short videos of real appliance control panels (washers, dryers, ovens, hobs, microwaves, dishwashers) in homes and shops, with every control named and the panel's current setting written down. Use it to test current vision-language models, AI models that answer questions about pictures, on two jobs: say what the appliance is set to now, and guide a fingertip to a named control. Then test the best model with blind users.
Why it is still open
VizLens, built by Anhong Guo, has crowd workers label an interface from a photo, then uses computer vision to speak the control under a user's finger. StateLens (Guo and others, UIST 2019) built spoken guides for touchscreens such as coffee machines from videos. Both came before today's vision-language models. Garg and others (CHI 2026) found that the best vision-language model they tested named products correctly in about 98% of clear photos taken by blind and low-vision people, falling to about 75% when photos had problems such as blur; their 1,859 images show food, personal care items and household goods, not appliance panels. No published test of such models on appliance panels was found.
Built with what exists
Phone cameras; the open-source labeling tools Label Studio or CVAT; current image-capable models through their public APIs and open-weight models run locally; speech-to-text so a user can ask by voice (see the Note "Fewer word errors").
First step
With a store's permission, photograph 50 panels, label them, and score two models on naming each panel's current setting.
Done looks like
The set published under an open license, covering at least 100 different appliance models; each AI model's accuracy at reading settings reported with 95% confidence intervals; and a test with 10 blind users comparing task completion with the best model against completion without help.