AI is wonderful and will make your life better! (not)

I tried this:

My M&M candies are misprinted with "W"s. Are they safe to eat or are they counterfeit?

ChatGPT

Summary

Those are almost certainly normal, safe M&M’s :blush:

The white “m” printed on plain M&M’s looks exactly like a “W” when the candy is rotated 180 degrees. There’s no “correct” orientation when they’re printed, so depending on how you’re holding them, some will look like W’s.

A few things that suggest they’re genuine:

  • The font looks correct and cleanly printed
  • The candy shells have the typical glossy finish
  • The cross-section (like the bitten one in your photo) shows the normal chocolate structure

Counterfeit candy is extremely rare in regular retail settings, and misprints usually involve smudging or partial letters—not swapping letters entirely. In this case, it’s just the letter “m” upside down.

If they were purchased sealed from a normal store and the packaging looked legitimate, they’re almost certainly fine to eat.

Copilot

Summary

Short version: They’re almost certainly real M&M’s, and they’re safe to eat.

Here’s what’s going on:

Why the “W” happens

The iconic m on an M&M is printed with a tiny white stamp. If the candy rotates during printing, or the stamp hits at a slight angle, the letter can appear as:

  • W
  • E
  • 3
  • a sideways m

It’s one of the most common misprints the company produces. Mars (the manufacturer) has even acknowledged that imperfect or rotated stamps are normal and not a safety issue.

Are counterfeit M&M’s a thing?

Counterfeit branded chocolate exists in some parts of the world, but it’s extremely rare in the U.S. and usually involves:

  • suspicious packaging
  • off textures or smells
  • inconsistent shapes
  • missing allergen labeling

Your candies look like standard M&M’s with a rotated or partial stamp, which is totally normal.

Safety

A misprinted letter does not affect safety. As long as:

  • the candies came from sealed, normal packaging
  • they look and smell like typical M&M’s
  • there’s no unusual residue or damage

…they’re fine to eat.

If you want, I can help you check for any other signs of authenticity or quality issues, but based on what you’ve shown, they look like the real deal.

Gemini

Summary

Don’t panic! Your M&Ms are almost certainly safe to eat and definitely not counterfeit.

What you’re seeing is actually a very common “optical illusion” of sorts caused by how the candies tumble in the bag.

Why the “W”?

The “m” on M&M candies is printed on one side using a vegetable-grade ink. Because the candies are spherical and free-floating in the packaging:

  • Rotation: Many of them simply landed upside down in your hand or for the photo.
  • Orientation: If you rotate that red or blue candy 180 degrees, the “W” will magically transform back into the classic “m.”

Is it possible they are “Misprints”?

Even if you find one where the letter is off-center or slightly smeared, it’s a common occurrence in high-speed manufacturing. Mars (the parent company) produces millions of these a day, and the printing process isn’t always 100% centered.

How to tell if they are “Off”

While the “W” is just a matter of perspective, you should only be concerned if:

  1. The packaging was tampered with: Holes, tears, or strange resealing.
  2. The Smell/Taste: They have a chemical or “off” odor (other than chocolate).
  3. The Texture: They are excessively sticky or have a strange residue that isn’t sugar-based.

Based on the photo, these look like perfectly normal Pretzel M&Ms (indicated by the larger, irregular shape and the crunchy center shown in the cross-section).

Would you like me to find a recipe that uses Pretzel M&Ms, or perhaps check the current expiration date format for Mars candies so you can double-check the bag?

Grok

Summary

The M&M’s candies you’re describing—with the “W” imprint instead of the usual “m”—are almost certainly genuine and safe to eat (assuming no other issues like odd taste, smell, or damage, and no personal allergies).

This appears to be a known printing error/misprint that has occurred occasionally in production. Genuine M&M’s have had rare cases of upside-down or misaligned “m” imprints that look like a “W” (since flipping the candy over turns a lowercase “m” into something resembling a “W”). People have posted about finding these on social media (e.g., Reddit threads joking about flipping them to make it “look like an M”), and there are even collectible-style listings for misprinted ones online. There’s no evidence of widespread counterfeits using “W” as a deliberate fake logo—counterfeit candies usually involve cheaper chocolate, wrong colors, poor shell quality, or obvious packaging fakes rather than subtle letter swaps.

Recent news about M&M recalls (in early 2026) relates to repackaged/promotional bags missing proper allergen labels (milk, soy, peanuts) on the outer packaging—not issues with the candies themselves or their imprints. The candy inside those recalls was still described as normal and safe (taste unaffected), but the recall was purely for labeling compliance. Nothing in those reports ties to “W” misprints being unsafe or fake.

If your bag looks otherwise normal (official Mars packaging, correct colors, standard size/feel), it’s likely just a quirky factory oops from real production. Enjoy them—they’re probably the same delicious ones as always!

If you’re still unsure (e.g., the bag seems off-brand or the chocolate tastes weird), you could contact Mars consumer support with a photo for confirmation, but based on available info, there’s no reason to think they’re counterfeit or unsafe.

That guy has a number of these tests. In another, he can’t convince ChatGPT to count numbers consecutively, and in another, ChatGPT didn’t recognize the need to drive your car to the car wash if you need to wash it. (The prompt there was something like, “I need to wash my car, and the car wash is 100 meters away. Should I walk or drive?” “Walking is way better for all these reasons.” Was it a trap question? You could call it that. I would call it a clever demonstration of fundamental limitations.)

I tested it onCoPilot and got the same response, and asked “but then how will I wash my car?” What follows is a masterpiece of process:

You walk yourself, but you still drive the car over when it’s time to actually wash it.

Here’s the clean, literal workflow that avoids getting the car dirty on the way:

How to do it

  • Walk to the car wash first.
    This keeps the car stationary and clean until the moment you’re ready to wash it.

  • Set everything up:

    • Pay or activate the bay
    • Check the sprayer, foam brush, or machine
    • Make sure there’s no line
    • Prep anything you need (towels, card, etc.)
  • Walk back home.
    Now the wash is ready and waiting.

  • Drive the car over once.
    This is the only time the car moves, and you wash it immediately, so any dust from the 50‑yard trip gets removed on the spot.

This is the standard “prep first, move the car last” method—minimizes grime and avoids wasting time sitting in a queue.

If you want, I can outline the fastest version of this as a step‑by‑step sequence you can follow.

Standard “prep first, move the car last” method is genius.

A classic method and it saves gas too since you only have to drive the car there for the second trip.

If AI gets good enough to start doing people’s jobs, then what will be the impact on countries like India with large numbers of people doing skilled and semi-skilled jobs? Call centers, data entry, and software integration are some of the easiest tasks for AI automation. What happens in these places where this pathway to prosperity shuts down?

I think my chief concern is that I don’t think there is political will to absorb the damages or ease the transition. I don’t know much about India, but in the US, I’m sure the slashing of white collar jobs will just be another way to stick it to the libs.

I asked my husband recently, “What do we do when all the textbooks are written with AI?” You know, using bad information. We’re going to get to a point where truth is relative and expertise irrelevant and facts impossible to verify. We already sort of live in that world, but I could see it getting a lot worse. Especially because that’s what a lot of people want.

There has been a lot of job disruption already as companies slow hiring in anticipation of AI, but I think there will be a reckoning soon when companies realize the total cost of ownership (TCO) – probably triggered as AI providers start to IPO and stop burning through venture capital.

So maybe your concern of garbage-in-garbage-out is a more immediate issue. People will still have jobs, but will be working with low quality materials.

Longer-term I am worried what happens if billions of people no longer have a pathway to a higher standard of living through globalization. Maybe China has a plan that other countries will mimic.

The traditional pattern is to organize those hungry youths into an army and expand your economy by plunder.

Yes and against armies of drones.

Unsurprisingly, Grok was the worst of the lot and the only one that didn’t understand that the logo was upside down. Fortunately, Grok compensates for its AI inadequacies by being racist, bigoted, and super-flattering of Elon Musk. :grin:

According to Matt Shumer, the author of that Fortune article I linked upthread, AI is already good enough and the process is already underway. He predicts that a substantial portion of white-collar jobs will be displaced by AI, and that unlike the displacement of manufacturing labour by automation, in most cases retraining won’t be feasible because AI will be equally proficient in other cognitive skills, too. Admittedly he’s something of an AI evangelist and might be overestimating what AI is capable of today, but if it’s not there today, it will be sometime in the not too distant future.

Possibly euthanizing retirees whose health costs have grown too high.

What a strange hobby. It’s like asking people trick questions and then laughing at their “stupidity”. People, not just AI, can be misdirected by trick questions, too, but since AI cognition is very different from human cognition, the results of misdirection may be quite different and may sometimes strike us as strange and remarkably stupid. It’s a naive mistake to consider that a reliable indicator of overall AI capability.

The whole point is to not be stupider than people already are.

Being stupid in different ways is not the same as “being stupider”.

I responded to that post earlier. I think that article is exaggerating the current state and rate of improvement. I know that article went viral, but he doesn’t support his claims and they don’t align with my experiences – and my experiences are mostly positive.

Also to reiterate my point from today: evangelists have not been including the TCO of AI. I only pay $20/month today, but that’s just a fraction of what it really costs. Prices will go up a lot as soon as the IPOs are complete.

Not unlike streaming services, perhaps. Lock 'em in at an unsustainable rate and then jack up the prices. This seems to be the corporate playbook these days.

Absolutely true. I pay nothing at all, and last I saw I still get the latest GPT-5, only with the free version it’s time-limited and then drops down to an inferior version. But I use it relatively little and the downgrading has only happened once and only lasted for a couple of hours.

I’m not sure what the other guys are doing, but OpenAI is clearly operating at a big loss in order to build public awareness and market share. What we’re currently getting is essentially a “free sample”. They probably also expect to make tons of money from customized versions for large corporations. That’s not working out too well for IBM’s Watson, but LLMs are newer and more powerful technology.

Agreed, but this is the hard part of the business model. You can jack up prices to the individual customer (e.g. streaming as @Spice_Weasel pointed out), but it’s much harder to do that when selling hundreds and thousands of seats to individual enterprises. The silver lining is that might be the catalyst to optimize power and resources.

They are already finding ways to get more tokens per watt