Ever since the days of the original Google Glass, I’ve had a mix of excitement and skepticism about smart glasses. With multiple promising products on the market, I decided to throw my hat into the ring to see if smart glasses could work for me. Given that I spend a lot of time around smart home gear and kitchen appliances, what better way to test smart glasses than to see how they hold up in the kitchen?
For this test, I’m using a sample of the Solos AirGo A5 Hydro 8 audio glasses — complete with prescription lenses because I’m useless without corrective eyewear. I approached this smart glasses cooking challenge from three angles:
- Basic cooking advice, like the science of food and identifying mystery produce
- Cooking a known recipe from a cookbook and finding appropriate dishes to complete a meal
- Learning a new recipe and checking it against the purported online source
Let’s see how they hold up.
About the Solos AirGo A5
Solos provided a sample pair with a prescription identical to that of my normal glasses.
Before I jump into each task, I would like to discuss the smart glasses I used and their capabilities. For this challenge, I used the Solos AirGo A5 smart glasses, which come with a companion app that essentially functions as an AI chatbot. The model I used doesn’t have a built-in display or camera; instead, it relies on the Solos app for those functions.
Originally, I planned to use the default chatbot instructions but it sometimes refused to help with cooking. Instead, I used Google Gemini to create chatbot instructions specific to cooking — with some flexibility to answer other questions. (While I have some experience with AI chatbot programming, fitting appropriate instructions into a character limit is a tall order for a verbose writer like myself.)
Unfortunately, the Solos prescription feels off for my left eye, so I’m unable to wear the glasses while working on my computer. Thankfully, the difference wasn’t an issue when cooking.
Solos has a few options for chatbots: GPT 4o Mini (Azure or OpenAI), Claude 3 Haiku and Gemini 2.0 Flash. (I settled on the Gemini bot because it’s the platform I’m most familiar with.) This means that my experiment is ultimately more about cooking with AI than with smart glasses — Solos (or any other brand with similar features) is merely a tool for speaking with a chatbot.
I found the voice synthesis in the Solos app a little robotic — at least compared to Gemini or ChatGPT. Still, it was easy to understand. Likewise, I appreciate having the option to adjust the speed, tone and response length, although the tone didn’t seem to affect the voice quality all that much in my experience.
Speaking of sound, Solos uses small speakers on the temples (the arms) of the frame but they are not headphones. This means that other people near you can also hear them when the volume is high. Solos has other features, such as live translation and fitness tracking, but these weren’t relevant to this experiment.
1. Basic cooking tips, conversions and identifying ingredients: Useful, but unstable
The cooking advice was generally accurate but it’s best to double-check your sources before taking anything at face value.
I didn’t find any deal-breaking issues with the basic cooking advice I requested in my tests. The smart glasses responded with helpful tips when I asked about boiling eggs or seasoning choices. It worked well for measurement conversions, like asking how many teaspoons are in 1/3 cup. Likewise, it accurately helped me convert a cup of cheese to grams, which accounts for volume and density.
After getting a quick recipe for a simple vinaigrette, I followed up with a question about why acidic foods taste as good as they do. While the information about acidity was useful and accurate, the chat model created article titles and links when I asked for sources to support its claims. This marks the first major hallucination of the experiment.
Next, I took a picture of some squash in my kitchen for identification. It accurately spotted a delicata squash, but had trouble with a larger squash that I later found out was a stripetti. At one point, it even told me the squash was a Korean melon, which wasn’t true.
The last test in this category was asking to triple a recipe in a photo from a cookbook, which only seemed to work when I specifically said, “I want to triple the size of this recipe.” Still, it was reliable enough on the few recipes I tried, including one with just four ingredients (alfredo sauce) and one with 11 (lasagna).
2. Cooking with a known recipe: Reliable, but quirky
Despite a few drawbacks, I believe that assistance with existing recipes is the best evidence that smart glasses have a place in the kitchen.
The biggest success of the experiment was the ability to take a photo of my recipe book and have the glasses walk me through the recipe.
I chose a simple Alfredo sauce from Betty Crocker’s Cookbook: Bridal Edition, which I make at least once a month. Not only did the smart glasses accurately capture all the ingredients and instructions, but they also helped me manage three recipes simultaneously for a complete meal. For example, I started the sauce before moving on to the chicken thighs and a simple side salad.
Transitioning between recipes was easy, usually involving a quick statement such as “I’m ready for the next step on the chicken” or “the sauce is now bubbling, what do I do now?”
After preparing all three components of the meal, I asked for advice on how to plate them. This made the final presentation look much better than my normal lazy serving style.
Still, I ran into two issues while cooking this meal: timers and fractions. Unfortunately, the AirGo A5 smart glasses don’t support setting timers and the double-tap method for triggering a Gemini timer on my phone is hard to master — hitting the specific part of the frame felt like a coin toss — it didn’t always work. Instead, I defaulted to my normal strategy of using the Google Nest Hub smart display in my kitchen.
As for fractions, the problem lies with the text-to-speech output, which ignores the slashes in phrases like “1/2 cup of cream” or “3/4 cup of shredded parmesan cheese” entirely. The result was a voice message calling for 34 cups of cheese. Oof. (In case you’re curious, scaling the entire recipe to accommodate 34 cups of cheese will create more than 181 servings.)
3. Learning a new recipe and checking it against an online source: Fail
To the credit of Solos, the first time it linked to Dash of Jazz was the original recipe, even if it hallucinated details and links in later messages.
Things went off the rails when I asked about learning a new recipe. I settled on jollof rice — a classic Nigerian dish that I love but haven’t ever cooked — it “provided” a list of ingredients supposedly from Dash of Jazz:
“Based on the Dash of Jazz recipe for Nigerian Jollof Rice, you’ll need the following ingredients: long grain rice, tomatoes, tomato paste, red bell pepper, scotch bonnet pepper (adjust to your spice preference), onion, garlic, ginger, curry powder, thyme, bay leaves, bouillon cubes, vegetable oil and salt.”
Let’s dig into this ingredient list to see where it goes wrong. The most notable entries are garlic and ginger, which aren’t part of the original recipe. Likewise, it omits tomato sauce and powdered white pepper from the list.
Because the whole purpose of the experiment was to follow the instructions on the glasses, I dropped by the store for ingredients I didn’t have on hand. When I followed up with Solos to obtain the quantities needed for the recipe, it continued to suggest garlic, while also adding cayenne pepper and smoked paprika to the list of ingredients that weren’t in the original recipe.
Upon further investigation, the follow-up recipe and quantities are nothing like jollof rice — and are also AI hallucinations. Fortunately, I had most of the fabricated ingredients, so the only extra I bought during my grocery run was 56 cents of fresh ginger.
Conclusion
I might change my tune with smart glasses that have a built-in display but even then, I’d rather stick to plain prescription glasses instead of a pricey gadget.
Cooking with smart glasses might be a great idea for following along with recipes but it’s not ideal for reliably finding new recipes or cooking techniques. This mostly stems from the reliance on AI chatbots — unless you’re willing to dig deeper for supporting sources.
As for the Solos AirGo A5, I like the glasses and the idea of having quick access to a chatbot for the simple stuff. The directional speakers are reasonably good, and having multiple chatbot models is a nice bonus. Still, tapping the on-device controls moves the frame just enough to give me motion sickness. (Voice controls are a decent alternative.) While the $250 price tag is affordable for smart glasses, the audio-only functionality might turn some people off.
Ultimately, I prefer using separate glasses, earbuds and the Gemini app for essentially the same functionality.
Read the full article here