Authorized adversarial testing for your next anti-cheat update.
Hire a cooperative adversary to build AI-assisted prototypes that reproduce player-level threats, test them using the same retail client as your players, and show your engineers exactly what happened.
RED-TEAM INSTINCTS.
PURPLE-TEAM OUTCOMES.
A new idea for fairer games Concept / 0.1
The next cheat can be built before the next player report.
We want players to win on skill. Community reports are valuable, but they start with what someone notices. They rarely come with the cheat’s code, intended behavior, failed attempts or a repeatable test.
Your next defense doesn’t have to wait for that report. BlindSpot puts a cooperative adversary to work after your anti-cheat update, while outside cheat developers are figuring out their next move.
A cooperative adversary you can direct, question and ask to retest. You know what was attempted and when, including the attempts your defenses stopped.
A player can buy a cheat, build one, or exploit the game without extra software. Our focus is the private, AI-assisted cheat developer who keeps changing their own tools.
Open a profile to explore the behavior and the coverage question.
Uses permitted controls and tools without deliberately exploiting the game. Skilled or unusual play can still look suspicious to someone watching.
Can the test distinguish unfair assistance from legitimate play? Clean control sessions give the studio a comparison point.
Finds an advantage inside the game’s own systems without needing an external program. A widely shared exploit may surface through community reports.
How much damage can the behavior cause, and can it be reproduced? A simple exploit can still have serious competitive or economy consequences.
Automates an action or a sequence of inputs. Some tools support legitimate accessibility; other uses provide a prohibited advantage. The game’s rules and actual behavior matter.
Can you identify the unfair behavior without misclassifying permitted use? Repetition alone does not make every automation case trivial to detect.
Needs little development skill to use a sophisticated product. Some users are blatant; others are restrained. Purchases give suppliers a reason to maintain their tools.
How does a distributed tool affect matches across its users and settings? Low effort for the buyer can still mean substantial harm for the game.
A player uses AI coding tools to prototype bespoke cheats, then revises them around their own play style. The tool may stay private, with no widely distributed version for your team to inspect.
Hire a cooperative adversary who reveals their code, intent and test sessions as they work. Private does not mean undetectable; it means your coverage needs testing against that specific behavior.
Has an incentive to restore a product after defenses change and keep paying customers. Teams may combine specialist engineering with distribution, support and ongoing maintenance.
How do you keep pressure on a supplier that keeps adapting? BlindSpot’s initial pilot targets player-level experiments; deeper specialist research needs its own scope.
A working model of access and behavior, not a universal danger score. Profiles overlap. Impact, reach and detectability depend on the game; the share of players building private AI-assisted cheats is not established here.
My background is web and software development, with very little game-development experience before this work. That outside perspective is useful: the pilot tests what a technically capable player without a game-development background can attempt with generative tools. Your engineers get documented results from that player-level starting point.
We choose the threat surface, such as information access, aim assistance or input automation, and the coverage question to test. Then we agree on the build, accounts, environment, permitted actions and stop conditions. You authorize the scope before work begins.
I use the same publicly available AI coding tools a vibe hacker can use, such as ChatGPT and Claude, to build scoped, player-level prototypes. I document what each version is meant to do, then test it on the agreed retail build.
My telemetry and working dossier update in near real time, so your engineers can watch results arrive, compare their logs and redirect the next test. You can also watch game sessions live through an optional private stream. Code changes, attempted behavior and session times stay in the record.
Build and test the small, private tools a player could make. Choose a focused subset of these examples around the behaviors your team wants to investigate.
See the evidence this producesYou get the prototype’s code, intended behavior and session evidence privately. Failed builds and blocked attempts stay in the record. A working bypass is never promised.
Outside cheat developers keep their work private. Your cooperative adversary shares the attempt, so your team can improve during the same post-patch blind spot.
Illustrative sequence, not a prediction of attacker timing or measured performance.
Switch on BlindSpot to see how your team can use the same window.
Drag the handle or tap a step to take control.
They reverse-engineer your update.
You get our debrief.
Inspect the kind of handover your engineers would receive.
Download the sample data
POST-PATCH DOSSIERProject Vanguard / v1.4.2The sample covers information advantage and aim assistance. Session observations show what the tester saw; your telemetry establishes what your defenses detected.
No enforcement was observed in selected sample sessions. Detection status remains unconfirmed until the studio correlates its logs.
Select a match to inspect its evidence.
Assistance disabled for the full session. Record normal input patterns and match context before comparing assisted behavior.
Client findings and working artifacts are handed over privately under the engagement terms.
The Post-Patch Pilot.
Hire a cooperative adversary to test your agreed retail build using AI-assisted prototypes. The work produces a private dossier your engineers can use throughout the engagement.
Concept offer for discussion. Pricing, delivery and guarantee terms would be agreed before any engagement.
Proposed guarantee: deliver the agreed dossier or waive the final 50%. Blocked attempts count as evidence. A successful bypass is never promised.
Creates a discussion brief with your selection. No booking or payment.
Controlled testing needs clear boundaries. The studio chooses the scope and can stop a session at any time.
The conversation that sparked the ideaBULKHEAD · THE HOTZONE, EP2No. It adds scoped adversarial testing and documented feedback to the systems and engineers you already have. No provider can promise that cheating disappears.
AI coding tools make custom development more accessible. A private tool may lack a known public signature, but behavioral and other detection layers can still catch it. We test a specific coverage question; we do not claim to know how many of your players build private cheats.
Keep listening to your players. This engagement adds a cooperative adversary you can direct: known test actions, private prototype code, controlled comparisons and planned retests. The studio can investigate an agreed threat without waiting for a player to encounter it and submit a report.
The proposed model fits an authorized beta or a released game’s update cycle. After launch, changes can create new questions about coverage. The pilot tests whether this approach produces useful evidence for your team; value is not assumed just because a game is live.
Only if the studio explicitly authorizes that environment and its rules. Controlled sessions are an alternative. Any approved live scope is unranked, capped and stoppable. No ranked ladders, economy abuse, server disruption or uncontrolled testing.
The proposed model starts with a retail client and no studio source access. Your team correlates our activity with its own telemetry. Account registration and enforcement settings are documented so results can be interpreted accurately.
This is Tim’s business concept, shared to gather feedback and potential collaborators. The pilot is a proposed offer. The sample report is fictional. No studio affiliation, customer result or available job is implied.
