A long LinkedIn post from a Microsoft engineer: "I have an open position for an IC5 Principal Software Engineer. My goal is to eliminate every line of C and C++ from Microsoft by 2030... Our strategy is to combine AI and algorithms to rewrite Microsoft's largest codebases... The purpose of this Principal Software Engineer role is to help us evolve and augment our infrastructure to enable translating Microsoft's largest C and C++ systems to Rust..."

There's a special irony in translating a codebase to Rust, for the main purpose of stability, using AI, which is well known to be unreliable in everything except confabulation. It's like protecting valuables by storing them in a vault made of swiss cheese

I can't WAIT to hear how this turns out

I dunno who needs to hear this but using AI at your Early Concept phase, or for ideation, is literally the worst place to deploy it

It isn't creative, what it does is hand back tropes. That's what the model does: sift for likely responses to an input

It's like building your foundation out of memes

Gen AI is trained to produce the most likely response, i.e. the kind of thing that appeared in its training set most often. If you're actually finding it useful for "idea generation", that just shows how creatively bankrupt you and your games are

Doing your budgets or schedules with a technology that routinely hallucinates is a terrible idea (as is contracts, legal advice, etc.).

Doing docs with AI is making the primary source of truth unreliable for the team and creates ambiguity or conflict, and thus also a terrible idea.

Festive cheer is coming to New Wirral! Enter the code HOLIDAYTREATS at the mailbox in Cassette Beasts to receive a rare wintery beast tape! 🎄

BYTTEN STUDIO: NEW PROJECT IN DEVELOPMENT MORE INFORMATION TO COME IN 2026

SPECIAL ANNOUNCEMENT: The team here at Bytten Studio has been working hard on our next game project, which we'll share more information on in 2026. We're really excited for it - we can't talk about it just yet, so please sit tight!

In the meantime, have a happy holidays from us! 🎃

A handheld purple and blue vpet with yellow buttons. On the screen is a Traffikrab - the traffic cone hermit crab from Cassette Beasts.

If you find this thread interesting, consider following me on mastodon, where I post more often - including about my custom vpet electronics project. mastodon.gamedev.place/@tccoxon

- items that widen the window of parries and are more likely to drop the lower your parry success rate is

- enemies that get stupider the worse you're doing, or even have a low chance to swap sides

More illusory difficulty mechanics come to mind now:

- a hidden luck stat that biases RNG in favour of the player, and increases each time they die. Karmic dice in BG3 are sort of like this (but I don't think high dice rolls make a game difficult unless the player can influence them)

Devs need to think about what experience they want the player to have, not "how difficult should this be"

Unlike parachute mechanics that catch the player as they fall, the goal with dynamic difficulty is usually to keep the player near a fall, but not fully going over. The nice thing about this is that when it works, in theory everyone gets a similar experience of difficulty regardless of skill/smarts.

Dynamic difficulty is usually a system that adjusts the encounters according to how well the player is doing. L4D is the best known example, but it can be done through self-correcting mechanics as well. “AI Director” isn't the only option.

I think of these as parachute mechanics because they catch the player as they fall.

I know that was all pretty vague, but there aren't many examples of the exact thing I'm talking about (that I know about at least). Check back in a couple of years.

But the players who do need it, receive the experience of surviving *by the skin of their teeth*, which is itself extremely rewarding. If the mechanic is perceived as a miraculous low-probability event, this doesn't even affect *the experience of difficulty* for the player who received assistance.

The goal would be for the player to not even know it's something that can happen until it's needed. Players doing well enough on their own don't receive the assistance, so it doesn't form part of their experience of the game's difficulty.

One thing you can do to engineer a less-demanding experience of difficulty is a sort of catch up mechanic that aids a player as they begin to fall behind, sort of like the items in Mario Kart, but far less overt.

With illusory difficulty, I mean that the difficulty *in the player’s experience of the game* can be higher than the *actual demand the game makes of the player*. Those are two completely different things - think about it.

Grind is just bigger numbers that demand more time from the player. It's a test of patience, not skill or strategy.

And then there are two more aspects I've been thinking about a lot lately: illusory & dynamic difficulty. The answer to difficulty isn't always to just add more sliders.

I debated whether to include grind here because I don't think it's really an aspect of difficulty, but here goes anyway: a lot of (particularly old) RPGs use grind to make their bosses feel bigger and badder in the absence of new (strategic difficulty) mechanics.

(Silksong and E33 are great games overall, even if they're not for me. Don't get me wrong, they deserve their praise.)

There's a lot more to difficulty than just punishment, skill, and strategy.

That was one of my frustrations with Expedition 33. I found the RPG elements (strategy) on their own easy, but I couldn't parry (skill) for shit. Except for some reason E33 has only one difficulty slider, so my only option was to make the already easy RPG even easier at the same time?

Like I said before, difficulty is multidimensional. Skill difficulty & strategic difficulty move along completely different axes. Players who are good at reflexes, rhythm, and coordination are not necessarily good at puzzles, planning, and technical understanding, and vice-versa.

Team Cherry’s stated goal with the difficulty was to encourage the player to explore alternative routes, but the corpse runs and traps actively work against that, so I speculate that they've mixed up difficulty and punishment.

Silksong isn't the only recent hit that IMO has misunderstood difficulty

Some N++ levels are absurd in their difficulty, especially when you're going for all the gold, but because losing costs nothing and restarting is instant, I stick with it, and I love it.

IMO, there wouldn't be half as much frustration at Silksong if it had just been less punishing.

Difficult games *don't have to be punishing*. In fact you can make games far more difficult when they're not, because players won't bounce off the game as quickly.

Look at N++. Despite my crappy reflexes and motor skills and the fact that it's a precision platformer, it's one of my favourite games.

Permadeath (in the absence of meta-progression, at least) is the ultimate punishment. But being set back to a checkpoint 15 minutes prior and made, through fear of losing yet more stuff, to replay the same segment of the game is also punishing.

Let's separate difficulty from punishment. If difficulty is the height of the hurdle, the punishment is how much it hurts your shins when you smack into it. It's the stuff that gets taken from the player, whether that's time or some in-game resource.

The game difficulty discourse is back and it's as tedious as ever. One thing that has become clear is that most people (even the devs of highly successful games) don’t seem to understand that difficulty has more than one dimension.

#gamedev

Promo graphic stating: Cassette Beasts has over 1.1 Million Units Sold (oh man, how did we manage that? Thank you all so much!)

We're so thrilled to be able to announce that Cassette Beasts has sold over *1.1 MILLION units*!

We started out as a tiny team hoping we might be lucky enough to sell even a fraction of that - for every share, review, or positive comment, we really can't thank all of you enough. 💜