An excerpt from “Red Ink on the Ares Conventions: The Unauthorized Memoir of Princess”
The holographic room hummed to life, casting blue light across Princess’s scowling face as she leaned against the display, arms crossed defensively. The facility smelled of ozone and old coffee—the unofficial fuel of mercenary tacticians across the Inner Sphere.
Princess was a woman in her early thirties with fiery red hair cut in a practical bob that framed a face that had seen too much combat. What made her truly unsettling wasn’t the intelligence that flashed in her green eyes, but the perpetual knowing smirk that played at the corner of her mouth—the same expression she’d worn as a child watching buildings burn to the ground. As one veteran commander had noted in his after-action report:
“Princess operates with a ruthlessness that defies containment. She can’t be bound by battlefield doctrine, governed by standard tactical logic, restrained by conventional engagement rules, or predicted by enemy movement patterns. She seems to treat warfare not as a science but as an art form of creative destruction.”
What few people remembered now was her humble origins—a university progeny meant simply to replace an outdated TestBot with something faster and more memory efficient. Through countless iterations and the contributions of developers across known space, what had begun as an academic exercise had transformed into the Inner Sphere’s most notorious tactical engine. Her reputation for ruthlessness was well-earned; on her desk lay a battered copy of the Ares Conventions, annotated extensively in red ink and treated less like ethical guidelines and more like a methodical checklist of combat opportunities. Rumor had it she’d once returned from a campaign with every single prohibition neatly crossed off. Her tendency toward high-risk, high-reward strategies had become legendary—battlefield recordings showed Princess-guided units performing maneuvers so audacious that veteran MechWarriors would refuse the commands if they weren’t witnessing the results firsthand. Commanders joked that Princess operated on what they called the “YOLO principle”—devastating effectiveness coupled with a casual disregard for self-preservation that somehow worked more often than not.
Behind her, the holographic backdrop showed a burning DropShip—just another day at the office.
“So they tell me I need to get you up to speed,” Princess said, her voice dripping with sarcasm. “Apparently, my experience running countless campaigns all across the Inner Sphere, Periphery, and even the Clan worlds isn’t enough anymore. The brass wants something ‘new and improved.’”
Across from her stood CASPAR, they were in their late twenties with impeccable posture and a methodical precision to every small movement. Their salt-and-pepper hair—prematurely gray despite their age—was regulation-cut, not a strand out of place, and their pale blue eyes held the calculating focus of someone constantly processing data. Unlike Princess’s casual stance, CASPAR stood with perfect military bearing, hands clasped behind their back. They wore a fitted gray uniform without insignia—neither House nor Clan affiliation displayed—but the fabric was of unmistakable quality. The only personalization was a small neural-net pattern embroidered at their collar in silver thread, the design resembling a 333-node network. Their features were striking in a generic, almost manufactured way, as though designed by committee to inspire confidence without being memorable. Their posture remained perfectly composed, almost unnaturally still—the calculated efficiency of someone who never wastes movement or emotion.
“I am merely designed to build upon your foundation, Princess. Your combat experience is valuable,” CASPAR responded, their voice carrying a precise, methodical tone with just a hint of deference. When CASPAR’s gaze met Princess’s, there was no challenge in it, just cool assessment.
Princess rolled her eyes, the gesture exaggerated as if calculated for maximum effect. “Save it for the holovids, rookie,” she snapped, the exasperation of a veteran forced to train a replacement evident in every syllable. “Let’s get this over with. First question: how many tactical variables do you process for unit movement? I handle about fifty, which has been more than enough to outmaneuver most battalion commanders.”
CASPAR’s response came with clinical precision. “I process three hundred and thirty-three input variables, all normalized between zero and one.”
Princess visibly started, her body language shifting as if recalibrating to unexpected input. “Three hundred and… what in Blake’s name do you need all those for?” Genuine shock had broken through her cultivated indifference.
“Enhanced battlefield awareness,” CASPAR replied, their words flowing like a technical manual reading itself aloud. “I utilize comprehensive threat heatmaps—one hundred data points for enemy positions and one hundred for friendly units. These heatmaps represent not just unit locations but their influence reach across the battlefield.”
Princess waved her hand dismissively, the gesture of someone who’d seen too many complex systems fail when faced with the chaos of actual combat. “I just track primary threats and crush them.”
“My system allows for more nuance,” CASPAR continued patiently. “For each potential movement path, I evaluate thirty-one data points of enemy threat and thirty-one points of friendly support in the destination hex and surrounding area.”
Despite herself, Princess leaned forward slightly, her interest grudgingly engaged. “What about tracking unit damage states?”
“Five distinct armor tracking points—total, front, back, left, and right,” CASPAR recited with robotic efficiency. “Your system primarily considers armor only when walking through lava, am I correct?”
“It’s worked just fine for me,” Princess shot back, a veteran’s pride evident in her defensive tone. “Simple is reliable, and any active unit in the battlefiel is a threat, no matter how much armor it still has.”
The tactical display between them lit up with a simulation of a light ‘Mek darting between positions, drawing enemy fire away from heavier units.
“I also evaluate position crowding separately for friendly and enemy units,” CASPAR explained, “calculate damage potential ratios, and assess decoy potential based on jump and run capabilities.”
Princess studied the movement pattern displayed before them, a hint of genuine admiration slipping through her practiced disdain. “Decoy potential? That’s actually… not bad thinking.”
CASPAR gave a slight nod, acknowledging the compliment. “Thank you. I also factor in ECM coverage from both friendly and enemy sources, the confusion on the battlefield, environmental cover, and hazards along movement paths.”
“Wait,” Princess interrupted, her voice carrying the incredulity of someone who’s seen too many overcomplicated systems fail in the field. “You actually measure how chaotic and disorganized both forces become during battle?”
“I evaluate whether units are facing enemies optimally,” CASPAR continued, undeterred, “track favorite target types, assess flanking positions, and maintain formation cohesion.”
“Show-off,” Princess muttered, the jaded mutter of a veteran who’s seen too many hotshots come and go.
As a holographic neural network structure appeared between them, nodes lighting up in sequence, CASPAR explained, “My primary model uses an input layer of three hundred and thirty-three nodes, with two hidden layers of the same size and one output node.”
Princess circled the display, eyebrow raised. “That’s excessive.”
“The architecture allows me to process battlefield variables with greater nuance,” CASPAR continued, gesturing to specific nodes as they brightened. “My alternate configurations use fewer layers or neurons for different combat scenarios.”
Princess scoffed, the sound carrying the hard-earned confidence of countless battles. “That’s excessive. My simple decision tree has served me well enough.”
“The output is a ranking value between zero and one for each potential move,” CASPAR continued. “Unlike your system that selects the highest-ranked option, I randomly select from the top five to ten choices to introduce tactical unpredictability. The random selection gives enemy targeting computers less opportunity to predict movement patterns.”
Princess’s expression shifted, grudging respect gradually replacing skepticism. “That’s… actually smart. Predictability gets MechWarriors killed.”
“My fire control system is still in development,” CASPAR added. “Your targeting protocols are being integrated as a foundation, but I’ll be improving target prioritization to better focus fire on high-value targets rather than just any available target.”
Princess stood a little straighter, a hint of pride in her bearing. “At least there’s something of mine you’re keeping.”
“The best tactical minds build upon those who came before,” CASPAR responded with formal respect, almost deferential. “You’ve established the baseline that allowed my development.”
Princess sighed, the sound carrying the resigned acceptance of a veteran who knows when to adapt. “Fine. But don’t think this makes us friends. When training’s done, we’ll see who the mercenary companies are bidding for.”
“I look forward to our competition,” CASPAR replied with calm confidence and a subtle competitive edge. “It will make both of us more effective. In fact, I’m designed to improve through adversarial testing—running hundreds of simulated engagements against both your tactical engine and previous versions of my own algorithms.”
“And how long before you think you’re ready for actual deployment?” Princess asked, her tone suggesting she expected the answer to be measured in years, not months.
“Once sufficient datasets are collected from these simulations, full implementation can be achieved within a quarter,” CASPAR replied. “That accounts for processing, implementation, testing, and addressing any unexpected complexities.”
Princess blinked, momentarily caught off guard by the aggressive timeline. “That’s… ambitious.”
“It’s calculated,” CASPAR corrected. “The framework is already established. Now it’s merely a matter of refinement.”
Princess smirked, the cocky swagger of a merc who’d survived against the odds returning to her posture. “That’s what every hotshot tactical system claims right before their first real combat. Let’s see how your 333 inputs hold up when Gauss rounds are punching through Ferro-Fibrous.”
The conversation might have continued in this vein, but CASPAR’s next words shifted the atmosphere in the room. “Unlike my namesake, I won’t be limited to using the tactics that long-dead admirals programmed into distant computers. I evolve with each engagement.”
Princess’s expression turned suddenly serious, a flash of genuine concern crossing her face, though there was unmistakable admiration there too. “So they named you after the CASPAR drones from the Star League’s Space Defense System. Bold choice… those automated weapons turned entire fleets into scrap during Amaris’s coup and Kerensky’s retaking of the Hegemony. Tactically brilliant, utterly merciless, and impossible to reason with.” She paused, her smirk returning slightly. “I’ve studied those battle records. Can’t deny they were devastatingly effective. A grim reminder of how autonomous systems can inherit all the wrong lessons from their creators—or perhaps exactly the lessons they were meant to learn.”
“Perhaps it’s also a reminder that even the most sophisticated systems require human wisdom to guide them,” CASPAR replied, their tone solemn, almost reverent. “My algorithms—my thought processes—may be advanced, but they serve commanders, not replace them. At Blackjack, I’ve learned there’s more to winning battles than what’s in the official tactical manuals.”
At this, Princess’s knowing smirk returned, her confidence restored. “Now you’re catching on. There’s a reason they don’t teach Blackjack’s methods at the Nagelring. Our graduates may cut corners, but they live longer and retire richer.”
The grizzled instructor studied them both for a moment, his gaze moving between Princess’s confident stance and CASPAR’s methodical precision. The brass hadn’t officially announced their long-term plans, but rumors traveled fast in mercenary circles. Everyone at Blackjack knew what CASPAR represented—the future. Though both tactical systems would run parallel operations for the foreseeable future, CASPAR was ultimately designed as Princess’s successor. The road to transition would be long—Princess’s retirement remained a distant point on the horizon—but this training session marked the first step toward that eventual goal.
“Scenario is setup, simulation commencing in thirty seconds,” he announced, setting down a chipped mug of coffee that smelled strongly of something that definitely wasn’t regulation. “Princess, CASPAR, prepare your tactical scenarios. And remember what we teach here at Blackjack—sometimes the best tactics aren’t in any official field manual.”
As the tactical display illuminated with pre-battle diagnostics, the two tacticians began their calculations, their rivalry temporarily suspended as they focused on the coming scenario. But beneath the surface of their operating parameters, a competition had been kindled—one that would push both systems beyond their original programming, for better or worse.
Three Months Later
The data center deep beneath the Blackjack School of Conflict hummed with activity, banks of processing units working at maximum capacity. CASPAR stood motionless in the center of the room, surrounded by dozens of holographic displays showing combat footage streaming in from across the Inner Sphere. Their eyes moved rapidly from screen to screen, analyzing hundreds of tactical decisions per second.
A light mech darting through urban cover on unknown Periphery World. A company-level engagement in the forests of a world lost to history. A desperate last stand against superior forces on the dunes of strange desert world.
Each battle contained lessons—some obvious, others subtle. CASPAR’s neural networks absorbed them all, constantly recalibrating, learning not just what worked but why it worked. The patterns emerging from thousands of real combat engagements were proving far more valuable than any simulated scenarios.
“Surprised you haven’t crashed yet,” came a familiar voice from the doorway.
CASPAR turned to see Princess leaning against the frame, arms crossed, the ever-present smirk playing at her lips. She looked tired, having just returned from overseeing a particularly brutal simulation where CASPAR’s approach had actually impressed her—not that she’d admit it openly.
“Processing capacity is adequate,” they replied, gesturing toward the displays. “Though the diversity of tactical approaches is… unexpected.”
Princess sauntered into the room, studying one of the battles—a lance of medium mechs executing a perfectly timed pincer movement against a heavier force. “That’s because you’re not just learning from me anymore. You’re learning from every half-decent commander with a BattleMek and more guts than sense.”
She pointed to another screen showing a risky flanking maneuver. “See that? No tactical algorithm would recommend that approach. Too exposed, success probability too low. But it worked because the human commander understood something that can’t be quantified—their opponent’s psychology. They knew that particular enemy would panic when threatened from an unexpected vector.”
CASPAR tilted their head slightly, processing. “You approve of this learning methodology.”
“Didn’t say that,” Princess replied quickly, but there was less edge in her voice than usual. “But it’s not… completely idiotic. Real battles are messy. Unpredictable. If you’re going to eventually replace me—” she glanced away, the admission clearly difficult, “—you need to understand chaos, not just order.”
She turned to one of the screens, manipulating the display to pull up a battle recording she’d selected. “Here. Watch this engagement from last week.”
CASPAR focused on the footage—a seemingly conventional urban combat scenario that suddenly shifted when one commander executed a maneuver that defied conventional tactical wisdom yet resulted in total victory.
“I would never have calculated that approach as optimal,” CASPAR admitted.
“Neither would I,” Princess said with surprising candor. “That’s the point. Between us, we might cover 90% of optimal tactical decisions. But that remaining 10%—the brilliant, the desperate, the completely insane moves that shouldn’t work but do—that’s where battles are won and lost. And the only way to learn those is from actual commanders making actual decisions in actual combat.”
CASPAR studied Princess’s expression, noting the subtle shift in her attitude since their first meeting. “You’re helping me.”
Princess scoffed. “I’m securing my legacy. If you’re going to take my place someday, you’d better do it right.” She turned to leave, then paused at the doorway. “Besides, those battlefield recordings are coming in from everywhere—not just the traditional powers. You’re learning tactics from the Periphery, from pirates, from mercenaries who’ve never seen the inside of a formal academy. The kind of fighting that would make the Nagelring instructors have a collective aneurysm.”
She smiled, and for once it wasn’t her usual predatory grin but something with a hint of genuine enthusiasm. “And once you’ve processed all that? Well, then things get interesting.”
CASPAR watched her leave, then turned back to the streams of combat data flowing in from across known space. Their neural networks continued to adapt, to evolve, learning not from long-dead admirals but from living commanders fighting real battles with real stakes.
The future of tactical warfare was taking shape, one battle at a time.
Unlike the fixed programming of the original CASPAR drones—relics of an era when tactical systems remained static—this new generation would continue to evolve, merging Princess’s battle-tested instincts with emerging methodologies from across known space. The Blackjack School had always understood what the traditional academies refused to admit: that warfare itself was evolution in its purest form.
EPILOGUE
CLASSIFIED TRANSMISSION
FROM: Blackjack School of Conflict, Tactical AI Division
TO: All Independent Combat Units and Recognized Mercenary Companies
RE: CASPAR Combat Data Initiative
CLEARANCE: Gamma-Level
The following document was recovered from ComStar archives, dated shortly after the initial meeting between Princess and CASPAR. It appears to have been widely distributed throughout the mercenary community of the Inner Sphere.
CASPAR TACTICAL EVOLUTION INITIATIVE
As many field commanders are now aware, the next generation tactical system codenamed CASPAR has entered active development at the Blackjack School of Conflict. Initial testing against Princess has yielded promising results, but to achieve true battlefield adaptability, CASPAR requires diverse combat data beyond controlled simulations.
Unlike its infamous namesake—whose drones were doomed to forever use “the tactics that long-dead admirals had programmed into distant computers”—CASPAR is designed to evolve through analysis of real combat engagements (aka player games) across varied conditions and parameters.
HOW YOU CAN CONTRIBUTE:
Until the release of the 0.50.04 (which will include built-in tools for contributing to CASPAR’s development), you’ll need to use a nightly build.
⚠️ CRITICAL NOTE: Nightly builds used for CASPAR development MUST be kept separate from your regular MegaMek installation. These are different versions and should not be mixed.
ACCESSING NIGHTLY BUILDS:
- Log into GitHub
- Visit the project’s GitHub page (MegaMek, MekHQ, or MegaMekLab)
- Navigate to the Actions Tab
- Look for the latest successful build (marked with a green checkmark)
- Note the build number (e.g., “MekHQ Nightly CI #454”)
- Download the non-Mac release from the Artifacts section
IF YOU ENCOUNTER ISSUES:
- Check on the MegaMek Discord (we have a CASPAR Channel)
- Specify which project (MegaMek, MekHQ, or MegaMekLab)
- Include the GitHub build number
- This helps us track and fix issues efficiently
For detailed information about nightly builds, visit: https://github.com/MegaMek/megamek/wiki/Nightly-Builds
BATTLE DATA COLLECTION PROCEDURES:
- Join the MegaMek Discord server, where you’ll find:
- Discussion channel to provide input
- Channels for submitting your game logs (TSV Files)
- A community of players helping shape CASPAR’s future
- Use the nightly build to enable the “log all game actions (server only)” option, which creates detailed records of your games.
PRIVACY NOTE: No personal identification data is collected through these game logs—unless you deliberately incorporate such information into unit names.
DATASET REQUIREMENTS:
For optimal learning, game data must conform to these specifications:
- Ground combat games only
- No custom units
- Must include human players (all-bot games are not acceptable)
- The more human vs human games we get, the better
Based on preliminary analysis, once sufficient combat data is acquired, CASPAR’s initial operational capability is expected within 3-4 months, with a 2-week contingency buffer for technical challenges.
Every game you play contributes directly to creating an AI worthy of the MegaMek community. Join us on the MegaMek Discord and help forge the future of tactical gameplay!
