At GDC 2024, King described AI as a way to test and refine Candy Crush Saga levels at scale—not as a replacement for level designers. The clearest reported uses were playtesting agents, level-quality assessment and tools to help designers iterate sooner. The account described workflow benefits, but did not publish figures for time saved, bot accuracy or effects on player retention or revenue.
What King presented at GDC 2024
In a talk reported by GamesBeat on March 21, 2024, Sahar Asadi, director of King’s AI Labs, and Anna Hernandelius, product director for Candy Crush Saga, discussed automated level creation and management, AI-assisted playtesting, and the challenge of maintaining quality as the game grows. GamesBeat reported that the game had more than 16,000 levels at the time; that is a dated figure, not a current level count. GamesBeat’s account of the GDC talk was updated June 17, 2025.
The underlying production problem is scale. A live game keeps adding levels, while each one needs checks for playability, pacing and difficulty. Manual review remains important, but becomes harder to apply alone across a large catalogue and a varied player base.
What “AI” meant in King’s account
The report describes several related uses, but they should not be collapsed into a claim that a generative model made the game’s levels. The most concrete work concerned agents that play levels and systems that help assess or refine them.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →#1 Best Overall
- Read Before You Buy — No Video Output: These adapters support charging and USB 2.0 data transfer, but cannot transmit video signals. Except for standard USB webcams (which use USB data only), they are not compatible with HDMI/DisplayPort cables, video-capable USB-C hubs, or docking stations with video output.
- Convert USB-A Ports to USB-C: Designed to connect USB-C earphones, cables, flash drives, card readers, and other USB-C accessories to standard USB-A ports. Plug-and-play with no drivers or software required.
- Aluminum Alloy Housing: Built with a sturdy aluminum alloy shell that aids in heat dissipation and protects against daily wear and scratches. Designed to maintain a stable and secure connection.
- Compact & Travel-Friendly: The ultra-compact design allows the adapter to stay plugged into your device without blocking adjacent ports or adding bulk, reducing wear and tear on your original USB ports.
- 12-Month Warranty: Backed by a 12-month manufacturer warranty for peace of mind. Designed to meet strict quality control standards for reliable everyday performance.
- Playtesting agents: Bots played levels before release to provide signals about how they performed.
- Quality and difficulty assessment: AI was intended to help flag levels that might be too easy, too difficult or frustrating. These are estimates to inform review, not objective measurements of fun.
- Designer iteration: Feedback from testing tools could help designers revise levels and spend less time on repetitive validation.
- Reinforcement learning: King discussed using reinforcement-learning techniques to develop agents intended to play more like people than like perfect solvers.
- Generative AI: King was exploring ways generative systems might assist designers. The report does not establish that generative AI independently produced shippable Candy Crush levels.
GamesBeat did not disclose the models, training data, internal architecture or specific generated outputs. The reported workflow is therefore best understood as AI-assisted evaluation and iteration, with generative assistance still less concrete.
Why a human-like agent is different from a solver
A bot optimized to complete a level efficiently can show that a solution exists, but that says little about whether an ordinary player will find the level clear, fair or enjoyable. King’s stated aim was to build agents whose play better reflects human behavior, so testing could offer more relevant signals about the experience players might encounter. “Human-like” describes the goal, not proof that the agents accurately reproduced the full range of real players.
Rank #2
- 5-in-1 USB-C Hub: Experience comprehensive connectivity featuring a Power Delivery input, two USB-A 2.0 ports, a USB-A 3.0 port, and an HDMI port. (Note: The USB-C power delivery input port is only for connecting an external wall charger to power your laptop and cannot power peripheral devices.)
- 90W Pass-Through Charging: Achieve optimal charging with 90W pass-through power to your laptop, supported by a total input of 100W, with the hub reserving 10W for operational efficiency. (Note: Wall charger not included.)
- Quick Data Transfers: Accelerate your productivity with rapid data transfers using a high-speed 5Gbps USB 3.0 port and two 480Mbps USB 2.0 ports.
- 4K HDMI Display: Enhance your visual experience with a hub capable of delivering 4K resolution at 30Hz in both mirror and extend modes. Please note that this hub is compatible with MacBook (macOS 12 and newer), Windows 10 and 11, ChromeOS, and laptops equipped with DP Alt Mode and Power Delivery. Note: This device is not compatible with Linux.
- What You Get: Anker USB-C Hub (5-in-1, 4K HDMI), welcome guide, 18-month warranty, and our friendly customer service.
| Approach | What it can reveal | Important limitation |
|---|---|---|
| Optimized solver | Whether a level appears theoretically solvable | Efficient play may not resemble ordinary play |
| Human-like agent | Potential difficulty, pacing and friction signals | Player behavior varies, and the report gives no accuracy measure |
| Human playtesters | Qualitative reactions, confusion and perceived fairness | Human testing is harder to repeat at high volume |
| Production telemetry | Behavior of actual players after release | It arrives after launch and requires interpretation |
The first two approaches are the ones directly discussed in the GamesBeat report. Human testing and telemetry are useful comparison points, not methods the report establishes as part of King’s disclosed system.
What results King reported—and what remains unproven
King’s reported benefits were operational: designers could get feedback earlier, iterate faster and spend less time on mundane checks. The tools were intended to help teams catch potentially poor experiences before players encountered them and to manage quality across a large, expanding level catalogue.
Rank #3
- Sleek 7-in-1 USB-C Hub: Features an HDMI port, two USB-A 3.0 ports, and a USB-C data port, each providing 5Gbps transfer speeds. It also includes a USB-C PD input port for charging up to 100W and dual SD and TF card slots, all in a compact design.
- Flawless 4K@60Hz Video with HDMI: Delivers exceptional clarity and smoothness with its 4K@60Hz HDMI port, making it ideal for high-definition presentations and entertainment. (Note: Only the HDMI port supports video projection; the USB-C port is for data transfer only.)
- Double Up on Efficiency: The two USB-A 3.0 ports and a USB-C port support a fast 5Gbps data rate, significantly boosting your transfer speeds and improving productivity.
- Fast and Reliable 85W Charging: Offers high-capacity, speedy charging for laptops up to 85W, so you spend less time tethered to an outlet and more time being productive.
- What You Get: Anker USB-C Hub (7-in-1), welcome guide, 18-month warranty, and our friendly customer service.
The public account did not give a percentage reduction in testing time, a comparison of bot predictions with human testers, a controlled study, or measured changes in satisfaction, retention, revenue or engagement. It also did not state how many levels were generated autonomously. In this context, “results” means workflow benefits described by King, not a publicly documented performance benchmark or independently verified business outcome.
Why human judgment remains central
Play data and simulated runs can help identify repeatable problems, but passing an automated check does not establish that a level feels fair, fresh or satisfying. A level can be solvable yet confusing, tedious or poorly suited to a particular group of players. Those judgments involve context—such as skill, accessibility, novelty and cultural expectations—that a single completion or difficulty signal cannot fully capture.
Rank #4
- Dual Converters, Infinite Potential：Includes 2× USB C male to USB A female adapters and 2× USB A male to USB C female adapters. Perfect for a wide range of uses—tablets with Bluetooth keyboards, expand USB ports on macbook, and more. Two different converters for all your daily needs
- Next-Level 10Gbps & 3A Charging: No more slow 480Mbps, this usb to usb c adapter has a transfer speed of up to 10Gbps, allowing you to do more transferring in less time. This usb adapter fits both USB A and USB C charger, supporting up to 3A fast charging
- Upgraded Exquisite Craftsmanship: With an aluminum alloy housing and metal connector, the usbc to usb adapter is extremely durable and sturdy. Rigorously tested to withstand more than 10,000 times of plugging and unplugging, ensuring long-lasting performance
- Broad Compatible: The usb c to usb adapter widely supports all USB C/ USB A devices like laptops, tablets, cellphones, car chargers, and phone chargers. Such as compatible with MacBook Pro/Air 2023/2022, Thunderbolt 4/3 Devices,Apple MagSafe Watch 9/8/7/SE/Ultra, iPad Pro 2022/2021, Samsung Galaxy S23/S20/S10, and iPhone 17/16/15 Pro. Plug and play
- Please Note: To reach 10Gbps speed, keep the cable under 3.3 ft. For USB A Male to USB C adapters, try flipping the USB C connector. USB C Male to USB A adapters support bidirectional 10Gbps transfer within 3.3 ft
This is an implication of the human-in-the-loop approach King described, not a quantified finding from the talk. Designers still need to interpret model feedback, decide what to change and approve the resulting level. That matters because a model can be confidently wrong, reflect a narrow range of player behavior or steer teams toward repetitive solutions if its recommendations are treated as decisions.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What other game teams can take from the case
King’s example is most relevant to studios with repeatable, high-volume content checks—not as a plug-in recipe for every game. Before investing in similar tooling, a team should be able to define what it wants the system to detect and how its predictions will be checked against people and production outcomes.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Best Value
- 5-in-1 Connectivity: Equipped with a 4K HDMI port, a 5 Gbps USB-C data port, two 5 Gbps USB-A ports, and a USB C 100W PD-IN port. Note: The USB C 100W PD-IN port supports only charging and does not support data transfer devices such as headphones or speakers.
- Powerful Pass-Through Charging: Supports up to 85W pass-through charging so you can power up your laptop while you use the hub. Note: Pass-through charging requires a charger (not included). Note: To achieve full power for iPad, we recommend using a 45W wall charger.
- Transfer Files in Seconds: Move files to and from your laptop at speeds of up to 5 Gbps via the USB-C and USB-A data ports. Note: The USB C 5Gbps Data port does not support video output.
- HD Display: Connect to the HDMI port to stream or mirror content to an external monitor in resolutions of up to 4K@30Hz. Note: The USB-C ports do not support video output.
- What You Get: Anker 332 USB-C Hub (5-in-1), welcome guide, our worry-free 18-month warranty, and friendly customer service.
- Start with a specific task. Decide whether the target is solvability, difficulty range, likely friction or another observable property. “Make levels fun” is not a sufficiently precise evaluation target by itself.
- Check scale and integration. Estimate whether testing volume justifies the system, and whether it can connect to the level editor, build pipeline and release workflow early enough to influence design.
- Represent different players. Test whether simulated behavior covers novice, typical and expert play, as well as accessibility-sensitive cases. A model trained on past behavior may miss new players or unusual strategies.
- Keep review and override with people. Designers need a way to understand why a level was flagged and to reject a recommendation that conflicts with the intended experience.
- Validate and maintain the system. Compare predictions with human feedback and, where appropriate, post-release behavior. Monitor for drift as mechanics, game balance or the player population changes, and govern any player data used in development.
These criteria follow from the practical limits of simulated play and the production challenges King described; they are not a claim that King disclosed a particular implementation for each one.
The organizational work behind deployment
Asadi emphasized that moving AI research into production requires collaboration and adoption by the teams making the game. That means researchers, technology groups, designers and production stakeholders have to work together: a prototype has little value if its feedback does not fit the team’s tools or arrive when a level can still be changed.
The report describes King building application-specific capabilities while also exploring AI assistance. It does not name outside vendors or identify a publicly available product. The lesson is not that a generic chatbot can reproduce King’s system; it is that useful AI depends on fitting a defined production problem and the people who need to act on its output.
What the public account leaves unanswered
GamesBeat’s report is an interview and conference account, not a technical paper or reproducible benchmark. It does not explain how closely agents matched real players, which measures defined level quality, how often human testers disagreed with the system, or how much time the tools saved. It also does not establish whether generative outputs shipped in the game or how player data was governed. Without those details, readers can assess the reported direction and workflow claims, but not independently measure the system’s accuracy or impact.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
