Dog in the Machine Makes Tactical Combat Feel Like Solving A Mechanical Puzzle

There is a particular kind of satisfaction that only strategy games seem capable of delivering. You stare at a battlefield long enough that what initially looked like a complete mess slowly starts making sense. One enemy is standing in the wrong place. Another is dangerously close to something explosive. There is a conveyor belt you ignored at first, a button you suddenly realize matters, and an ability in your hand that might connect all of those pieces together. A few minutes later, the plan works and half the battlefield collapses in one ridiculous chain reaction.

Dog in the Machine seems built almost entirely around chasing that feeling.

Developed by French indie team Studio Chahut and published by Goblinz Publishing, Dog in the Machine is a tactical turn based roguelite arriving on Steam on September 29. You play as an AutoDog attempting to escape a dystopian mecha city and reunite with the humans who were exiled from it. That is the larger goal, but getting there means surviving a series of compact tactical battles where almost everything in the environment can become part of the solution.

The battlefields are small enough that every object feels intentional. Conveyor belts can move things around the grid, explosive obstacles can turn a basic attack into something much larger, enemies can collide with one another, and interactive mechanisms can completely change the state of a fight. It does not look like the kind of strategy game where you simply identify the strongest enemy and dump damage into it until the problem disappears. Instead, you are constantly looking for ways to make the battlefield hurt itself.

That is probably the biggest reason Dog in the Machine stands out to me. Turn based combat can sometimes fall into a familiar rhythm where the player eventually discovers the efficient sequence and repeats it. Here, the environment appears to be just as important as whatever abilities you have equipped. A Trick that seemed fairly ordinary in one battle might suddenly become devastating because an enemy is standing beside the right hazard in the next one.

The game calls the AutoDog’s abilities Tricks, which is exactly the kind of terminology I would expect from a game starring a robotic dog. After each fight, the AutoDog learns something new, but the real depth comes from combining those Tricks together. Two existing abilities can be merged into a more complex one, either by following available blueprints or experimenting until you create something that fits the way you want to play.

That immediately opens the door for the kind of broken builds that make roguelites fun.

There is always a point in a good deckbuilder or roguelite where the run stops feeling like the game gave you a collection of random upgrades and starts feeling like you accidentally invented something terrible. Two abilities begin working together better than expected, then another upgrade amplifies the combination, and suddenly the build has gone from functional to absurd. Dog in the Machine appears to actively encourage players to look for those interactions rather than steering everyone toward the same handful of obvious choices.

The battlefield adds another layer to that experimentation because the strength of a Trick depends on what you can do with it in the space around you. An attack that pushes an enemy may not look particularly impressive until pushing that enemy causes a collision, sends another unit onto a conveyor belt, and eventually triggers an explosion. The satisfaction does not come from seeing a huge damage number. It comes from realizing that you created the sequence.

You Can See Your Mistakes Before They Happen

The feature I suspect will make or break Dog in the Machine is its preview system.

Before committing to a sequence of actions, players can see how their instructions are expected to play out. If the result is awful, you can cancel everything, change the plan, and try again. You are not punished because you misunderstood whether an enemy would move one tile or two, or because an environmental interaction behaved slightly differently than expected.

That may sound like the game is making tactical combat easier, but I think it could actually make the strategy more interesting.

Some tactical games train players to become overly cautious because experimentation is expensive. If trying something unusual could ruin the entire run, the safest option becomes increasingly attractive. You stop wondering whether the ridiculous plan might work because you do not want to deal with what happens if it does not.

Dog in the Machine gives you permission to find out.

You can arrange a sequence just to see what happens, realize it creates an even bigger problem, and immediately start rebuilding it. The challenge becomes finding the best solution rather than trying to predict hidden information.

That feels particularly appropriate for a game where so many environmental elements can interact at once. If I am trying to calculate a chain involving three enemies, a conveyor belt, an explosive object, and two custom abilities, I would much rather spend my time solving the actual puzzle than restarting because I misread one interaction.

The developers describe the philosophy as tactics without the headache, and for once that description seems pretty accurate. The systems can still be complicated. The player simply gets enough information to understand what those systems are doing.

If you lose, the idea is that your strategy failed.

That is a much more satisfying reason to lose.

Building The AutoDog Becomes Its Own Puzzle

The battles may be where all of those systems come together, but the spaces between them appear just as important to a run.

As the AutoDog moves through the city, it encounters different characters who reveal more of the larger story while offering opportunities to expand the Trick Library. New abilities can be bought, exchanged, upgraded, looted, or stolen, meaning that progression is tied directly into the people and situations you encounter rather than being handled exclusively through a post battle menu.

Those choices should gradually push each run in a different direction.

Maybe you begin with abilities that encourage direct attacks but eventually discover a combination built almost entirely around moving enemies. Another run could revolve around explosive effects. A third might turn into something completely different because you found a Trick that seemed unimportant at first and then realized it paired perfectly with something already in the deck.

The more interesting roguelites rarely give you the exact build you wanted when you started. They give you pieces and force you to figure out what kind of build those pieces can become.

Dog in the Machine appears to understand that.

I also like that the game has an actual motivation beyond escaping an endless sequence of tactical rooms. The AutoDog is trying to find its humans after they have been exiled from the city, and the characters encountered along the way gradually reveal what happened to this strange mechanical world.

There is something unexpectedly charming about that premise. You have this machine learning increasingly complicated Tricks and becoming better at dismantling entire battlefields, but underneath all of that tactical optimization is a dog trying to get back to its people.

That gives the game just enough emotional weight to keep the run from feeling completely mechanical.

Small Battlefields May Be The Smartest Choice

One thing I appreciate about Dog in the Machine is that it does not seem interested in pretending bigger automatically means deeper.

The grids are deliberately compact.

That matters because smaller battlefields make every piece of the environment more important. A single conveyor belt can influence the entire encounter. One explosive object may completely change where you want enemies positioned. A button on the opposite side of the board may suddenly become more valuable than attacking the enemy standing directly in front of you.

On a massive tactical map, those individual elements could easily become background noise. Here, everything has the opportunity to become part of the plan.

It also means players can focus more on interactions and less on moving units across empty space. Dog in the Machine does not appear to be interested in simulating a huge military conflict. It wants to give you a small mechanical puzzle and enough tools to completely dismantle it.

That sounds far more appealing to me.

The best moments in this kind of game are rarely the ones where you simply overpower the enemy. They are the moments when you notice something the game never explicitly told you to do. You use an ability in a strange way, combine two mechanics that probably should not work together, and suddenly the encounter is over before the enemy ever gets the chance to execute its own plan.

Dog in the Machine looks like it wants those moments to happen constantly.

Studio Chahut says its goal is to make intellectual challenges enjoyable for a wider audience, and this feels like a smart way to approach that. The game is not stripping away strategic depth to make itself approachable. Instead, it is giving players better tools to understand that depth.

There is a major difference.

Anyone can make a tactical game difficult by hiding information or punishing mistakes. Making players feel clever when they finally see how all the pieces connect is much harder.

Dog in the Machine launches on Steam on September 29, and if its preview system, custom Tricks, and reactive environments work together as well as they sound, this could be one of those strategy games where staring at the board for five minutes is just as entertaining as watching everything explode afterward.