This is a coordinated, global effort to address what is arguably the single most significant bottleneck in the race to Artificial General Intelligence (AGI): the fundamental algorithms used for training neural networks.

The Highest-Leverage Point in the AI Stack

Progress in AI is increasingly driven by algorithmic innovation, not just the brute-force scaling of compute. At the very centre of this dynamic is the optimiser: the algorithm that dictates how efficiently a neural network learns.

The field still leans almost entirely on a single innovation from 2014: the Adam optimiser. With over 220,000 citations, Adam is recognised as the most influential AI paper in history. It enabled the stable training of the Transformer architecture, unlocking hundreds of billions of dollars in value. It is the foundational engine of the modern AI world. And it is a decade old.

The reliance on this decade-old algorithm represents the highest-leverage opportunity in the entire AI stack. The impact of a superior optimiser is not linear; it compounds across the entire global R&D landscape.

A breakthrough could save tens of billions of dollars in training costs globally. But the impact goes far beyond cost, a qualitatively better optimiser could enable the training of entirely new types of AI models with emergent capabilities required to achieve AGI.

The Limits of Centralised R&D

If the optimiser is so crucial, why does the field still rely on Adam? The answer lies in the limitations of the current R&D paradigm.

The conventional approach (proprietary research within centralised labs) is hitting its limits. The immense costs of training frontier models mean that only a handful of well-funded labs can iterate at the cutting edge. This centralised model drastically slows the global pace of discovery, excludes the vast majority of the world’s scientific talent, and is proving too slow and too narrow to solve this fundamental problem.

The TIG Paradigm: The Decisive Advantage of Openness

We need a paradigm shift in how foundational algorithms are developed. History proves that, given the right incentives, closed systems cannot compete with open, competitive ecosystems.

Linux dominates global infrastructure, and Bitcoin commands unparalleled compute power. They achieve this not despite being open, but because they are open. Similarly, open science inherently outcompetes closed science because researchers are free to experiment and build immediately upon one another’s work. The scientific feedback loop is accelerated. Closed, proprietary systems simply cannot match this level of parallel experimentation and compounding improvement.

The Innovation Game is purpose-built to harness this dynamic. TIG provides the incentive structure: a “synthetic market” (as detailed in the TIG white paper) that makes open, collaborative development the competitively superior model for the algorithms that underpin science, technology, and AI.

The Optimiser Challenge: Activating the Market

The Optimiser Challenge will leverage the power of TIG’s synthetic market, which will be specifically aimed at discovering the ‘next Adam’.

TIG provides the platform and the economic incentives for a global community of researchers (“Innovators”) to relentlessly collaborate and compete. Their optimisation methods are stress-tested in real-time by a decentralised network (“Benchmarkers”), who are economically incentivised to identify and adopt the most efficient solutions. This creates immediate demand and price discovery for the best algorithms.

This structure ensures that the best ideas rapidly surface, are built upon, and compound. It fundamentally shifts AI development from an arms race of capital, and held back by secrecy, to a rapidly evolving global contest of ideas.

The Mission

We are launching an effort to build a new engine for AI that is faster, more efficient, and guaranteed to remain open.

The race to replace Adam has begun, and it will be won in the open. The challenge is now live on testnet, with the mainnet launch just weeks away. We invite the global community to join this crucial effort:

1. Innovators (Researchers & Developers): Access the challenge documentation and start developing your optimization methods on the TIG testnet http://docs.tig.foundation/innovating.
2. Benchmarkers (Compute Providers): Prepare your compute resources to validate submissions and earn rewards. Learn more here
   https://docs.tig.foundation/benchmarking.
3. Forum Discussion: Join the conversation and contribute to the ecosystem on our Forum
   https://forum.tig.foundation/t/upcoming-challenge-neural-network-gradient-descent/47.

By creating an ecosystem where the best ideas (not the biggest budgets) win, we accelerate the path to AGI and ensure the next great breakthrough belongs to all of us.

The TIG protocol was founded on a bold hypothesis: that a competitive, incentive-driven ecosystem could significantly accelerate the development of highly performant algorithms for complex computational problems. We are thrilled to announce that this hypothesis has been validated. The top-performing algorithm developed within the TIG ecosystem for the Quadratic Knapsack Problem (QKP) now rivals and, in some respects, surpasses established state-of-the-art (SOTA) methods, marking a pivotal proof-of-concept for our platform.

This achievement is more than just a new solution method to a difficult problem; it is a demonstration of TIG’s power to cultivate and accelerate innovation. By transforming complex challenges into competitive opportunities, our “synthetic market” has successfully driven the creation of an algorithm that has advanced the state of the art in the field.

The Challenge: The Quadratic Knapsack Problem

The Quadratic Knapsack Problem is a notoriously difficult combinatorial optimization problem with significant real-world applications, from logistical planning and financial modeling to telecommunications and resource allocation. Finding an optimal solution is computationally intensive, pushing the boundaries of algorithmic efficiency. It is for this reason that we chose the QKP as a benchmark challenge for the TIG protocol — if we could stimulate progress here, we could do it anywhere.

The Journey to State-of-the-Art

The accompanying figures demonstrate the remarkable journey of innovation within the TIG ecosystem. The first figure illustrates the evolution of solution quality using the Relative Percentage Deviation (RPD) from the optimal solution, where lower values indicate better performance. The second figure highlights algorithm efficiency by plotting the total computational time spent searching for the optimal solution across all benchmark instances. Together, these figures capture the iterative advancement and progressive refinement of algorithms over rounds in TIG.

Our comparison begins from Round 44, marking the evolution of the Knapsack Challenge into the more complex and ambitious Quadratic Knapsack Challenge. Early submissions provided foundational solutions. However, a clear and rapid progression followed quantified by a decrease in runtime coupled with an improvement in solution quality, with the RPD value dropping by orders of magnitude.

This iterative enhancement illustrates the core strength of the TIG model. Because all algorithms are openly available and visible, participants can directly build upon previous solution methods, continuously refining their approaches. Our protocol incentivizes not only groundbreaking ideas but also the incremental yet crucial optimizations in speed and memory usage that are vital for real-world applications. As innovators compete and collaborate openly, they push each other forward, progressively improving both solution quality and algorithm performance.

Our current top-earning algorithm finds high-quality solutions in remarkably fast runtimes. This submission competes on solution quality with the SOTA methods of GRASP+Tabu¹ and IHEA², and notably exceeds the performance of the DP+FE³ and QKBP⁴ algorithms. Most significantly, it achieves this solution quality in runtimes that are orders of magnitude faster than existing SOTA methods. Only on the LargeQKP⁴ benchmark does the QKBP⁴ algorithm have slightly faster runtimes but this is coupled with lower solution quality. Older SOTA algorithms like DP+FE³ and GRASP+Tabu¹ were not evaluated on larger benchmarks of QKPGroupIII⁴ and LargeQKP⁴ due to excessive runtimes.

In summary, our current top-performing TIG algorithm delivers high solution quality with exceptional runtime efficiency, a critical advantage for industries needing to process vast datasets under tight time and cost constraints.

Why This Milestone Matters

1. Proof of Concept: TIG’s model for open innovation is no longer theoretical; it is a proven engine for algorithmic advancement. We hypothesized that by creating a competitive environment with tangible incentives, we could accelerate the development of superior algorithms. The success in the QKP challenge validates this vision.

2. Speed and Efficiency as a Driving Force: The TIG competition structure inherently rewards efficiency. When solution qualities are comparable, speed and lean memory usage become the deciding factors. This dynamic has pushed innovators to develop algorithms that are not only accurate but also incredibly fast, a feature often overlooked in purely academic pursuits but critical for industrial viability.

3. A New Paradigm for Innovation: This result demonstrates that TIG can systematically drive progress on established SOTA benchmarks. By iteratively pushing for better solutions, our platform incentivizes the kind of focused, competitive refinement that leads to world-class results.

The Next Frontier

Having proven the success of the TIG protocol on a problem as fundamental as the QKP, we are now setting our sights on new horizons. The next phase of our mission is twofold: deploying this powerful innovation engine to address critical challenges in important industry-focused domains, and pioneering a new AI-augmented approach to accelerate algorithmic innovation, positioning TIG at the forefront of algorithmic discovery.

By launching new challenges across diverse fields such as artificial intelligence, technology, and scientific research, we will leverage the TIG ecosystem to generate licensable, SOTA algorithms. This will not only create significant value and generate revenue for the protocol but will also provide industries with the high-performance computational tools needed to solve their most pressing problems. The journey with the Knapsack Challenge has been a remarkable success, and it is only the beginning.

References:

[1]: Yang, Z., Wang, G., & Chu, F. (2013). An effective grasp and tabu search for the 0–1 quadratic knapsack problem (GRASP+tabu). Computers & Operations Research, 40(5), 1176–1185.

[2]: Chen, Y., & Hao, J. K. (2017). An iterated “hyperplane exploration” approach for the quadratic knapsack problem (IHEA). Computers & Operations Research, 77, 226–239.

[3]: Fomeni, F. D., & Letchford, A. N. (2014). A dynamic programming heuristic for the quadratic knapsack problem (DP+FE). INFORMS Journal on Computing, 26(1), 173–182.

[4]: Hochbaum, D. S., Baumann, P., Goldschmidt, O., & Zhang, Y. (2025). A fast and effective breakpoints heuristic algorithm for the quadratic knapsack problem (QKBP). European Journal of Operational Research, 323(2), 425–440.

TIG Algorithms: https://github.com/tig-foundation/tig-monorepo/tree/main/tig-algorithms/src/knapsack

Introducing the Balanced Hypergraph Partitioning Challenge — a pivotal step towards faster, more efficient, and open artificial intelligence.

By The Innovation Game | Aoibheann Murray

The race to build more powerful AI is on. We see it every day in the headlines: larger models, new capabilities, and ever-increasing computational demands. But behind the scenes, this explosive growth is pushing against a fundamental wall. The very systems we use to train these massive models are becoming a bottleneck.

At The Innovation Game, we believe the algorithms to these foundational problems should be developed openly. That’s why we’re proud to launch our most ambitious challenge yet: Balanced Hypergraph Partitioning.

This challenge tackles a critical efficiency problem that impacts everything from the design of next-generation microchips to the training of large-scale AI. By focusing on GPU-accelerated parallel algorithms, we are creating a platform for the global community to collaboratively engineer the future of high-performance computing.

The Genesis of the Challenge

Every great idea has an origin story. Ours begins with Professor William Knottenbelt of Imperial College London, co-author of Parkway, one of the first scalable parallel hypergraph partitioners. He highlighted how hypergraph partitioning addresses one of modern computing’s critical bottlenecks — a problem that, if solved, could unlock enormous potential for distributed AI training and decentralized technologies.

Professor Knottenbelt and his students recently explored the idea of developing an automated tool designed to enhance hypergraph partitioning algorithms. Inspired by the success of tools like Google’s AlphaEvolve, which used AI to discover new algorithms, we envisioned creating an open, automated “algorithm innovator” accessible to everyone on our platform. While realizing this tool might take considerable time, the prospect underscored the importance of including hypergraph partitioning as a central challenge for TIG.

By designing a challenge that requires GPU-specific parallel algorithms, we can attract top researchers and frame the problem for this new paradigm. Hypergraph partitioning was the clear choice to kickstart this new chapter for The Innovation Game.

So, What Exactly Is Hypergraph Partitioning?

To understand the challenge, you need to understand the structure at its heart: the hypergraph.

You’re probably familiar with a standard graph, which has nodes (points) connected by edges (lines). An edge connects exactly two nodes. But what if you need to represent a more complex relationship, like a task that requires say 100, or even 1000 different processors?

That’s where a hypergraph comes in.

• Nodes: These represent individual things — GPUs in a cluster, components on a circuit board, or data points in a machine learning model.
• Hyperedges: Unlike regular edges, a hyperedge can connect any number of nodes at once. It represents a shared relationship — a task, a common property, or a dependency.

Hypergraph partitioning, then, is the problem of dividing all the nodes into a set number of groups, with two crucial goals:

1. Balance: Each group should have a roughly equal number of nodes (or weight).
2. Minimize Cuts: You want as few hyperedges spread across groups as possible.

In simple terms, it’s about putting related things together to reduce “chatter” between groups. For distributed AI, this means less communication between GPU clusters, which translates directly to lower latency, faster model training, and more efficient use of power.

A Challenge Ripe for Breakthroughs

Hypergraph Partitioning is what’s known as an “NP-hard” problem. This means there’s no easy, brute-force way to find the perfect solution, especially as the scale increases. Progress depends entirely on clever heuristics and novel algorithms. This combination of inherent difficulty and real-world importance makes hypergraph partitioning an especially compelling and impactful area of research.

Although the problem itself isn’t new, its significance has recently increased considerably due to potential wide-reaching impacts across various high-value industries:

• High-Performance Computing (HPC): Partitions massive simulations running on supercomputers to minimize communication between processors and maximize efficiency.
• AI and Machine Learning: Accelerates the training of large-scale models by intelligently distributing the workload across GPUs.
• VLSI Circuit Design: Optimizes the physical layout of computer chips to reduce wiring, minimize power consumption, and decrease latency.
• Distributed Systems: Balances workloads in cloud databases and decentralized networks to ensure smooth, fast performance.

Despite decades of research, it is believed that there is still much to be discovered in the field. Recent discoveries prove that significant leaps are still possible, and any novel approach could lead to measurable performance gains and valuable intellectual property. It’s a challenge that is hard enough to be interesting, valuable enough to matter, and perfectly timed for today’s AI-driven search methods.

Our Vision for Open AI Advancement

We believe the AI revolution should not happen behind closed doors. As our co-founder, Dr. John Fletcher, recently argued, keeping AI innovation open is critical for ensuring a future where its benefits are shared by all.

Our Balanced Hypergraph Partitioning Challenge marks a pivotal first step in this mission. This is a call to action for the global community to collaboratively tackle the fundamental bottlenecks limiting AI’s potential, starting with this challenge and continuing with future initiatives like Neural Network Training Optimization. The groundbreaking algorithms developed will remain open and accessible, paving the way for a more powerful and equitable technological future.

Acknowledgements

We’d like to thank Dr. Tasuku Soma for his expert guidance in the development of this challenge. Dr. Soma is an Associate Professor at the Institute of Statistical Mathematics and the Graduate University for Advanced Studies, SOKENDAI, Tokyo, and a former fellow at the Department of Mathematics at MIT, where he worked under the mentorship of Michel Goemans. His research expertise lies at the intersection of combinatorial optimization and machine learning, with a focus on submodular optimization and linear algebra in combinatorial optimization and algorithm design. We look forward to continued consultation with Dr. Soma as this challenge evolves and develops further.

At The Innovation Game (TIG), our core objective is to incentivise coordinated open development of algorithms. We achieve this by creating a platform where algorithm submissions are added to an open whitelist, and executions can be independently verified through a deterministic runtime signature. This allows execution of private algorithms to be detected and penalized, letting TIG enforce openness — a foundational integrity requirement for our ecosystem.

Initially, we built this atop WebAssembly (WASM) for its sandboxing and portability. However, as our ambitions and the complexity of real-world challenges grew, WASM became an obstacle. This article outlines why we transitioned to a high-performance native runtime, how we preserved critical safeguards like runtime signatures and fuel metering, and where we’re heading next.

Why We Moved Beyond WASM

Our original model compiled submitted algorithms into WASM and executed them in a custom virtual machine (VM). This gave us tight control: we could compute runtime signatures for verification, assign execution costs (fuel), and measure memory usage.

But WASM introduced growing limitations:

• Performance Bottlenecks: The VM layer added substantial overhead. Coupled with our compute budget constraints, this limited the size and realism of challenge instances.
• No Hardware Acceleration: WASM couldn’t tap into modern CPU extensions (like SIMD) or run on accelerated hardware like GPUs. Cutting-edge algorithm performance remained out of reach.
• Single-threaded Execution: WASM’s threading model made it difficult to scale across multiple CPU cores, limiting throughput.
• Inefficient Memory Model: WASM’s linear memory made allocations slow and brittle compared to native systems using virtual memory.

We needed to break out of the WASM sandbox — without sacrificing the guardrails it offered.

A Native Runtime for x86–64, AARCH64, and CUDA

We designed and implemented a new native runtime to execute algorithms as compiled machine code on:

• x86–64 — Intel & AMD CPUs
• AARCH64 — ARM CPUs
• CUDA — Nvidia GPUs

The major challenge was keeping our safety mechanisms — particularly runtime signatures and fuel limits — without relying on a VM. Instead of trapping execution in a sandbox, we instrument the compiled code with safety checks.

Runtime Signature: Verifying What Actually Ran

To enforce algorithm transparency, every run of a whitelisted algorithm must produce the same runtime signature — a fingerprint of the exact execution path. Here’s how we compute it:

On CPU (x86–64 / AARCH64)

We use a custom LLVM pass to instrument code during compilation:

• Every instruction is assigned a unique prime-based ID, modified by its location.
• As the algorithm executes, these are mixed into a per-thread signature.
• At function exits, signatures are flushed from thread-local storage into a global accumulator.

This ensures the signature remains consistent, even across multi-threaded runs.

On GPU (CUDA)

We parse the low-level PTX assembly post nvcc compilation and inject code into every kernel:

• Each basic block updates a signature register with a unique constant.
• The final signature is written to global memory for host-side verification.

This preserves determinism while respecting the GPU’s massively parallel nature.

Fuel Metering: Controlling Runaway Computations

To prevent denial-of-service risks (e.g. infinite loops), we enforce a fuel limit on all algorithm executions.

On CPU (x86–64/AARCH64)

• Each instruction has an associated fuel cost.
• Fuel usage is tracked in thread-local storage and periodically committed to a global counter.
• If fuel is exhausted, the program terminates via a trap.

On GPU (CUDA)

• Fuel usage is accumulated in registers during kernel execution.
• A check at kernel return compares against a fuel limit (injected at runtime).
• For finer-grained control, developers can insert CHECK_FUEL_LIMIT macros in their CUDA code, which we transform into custom fuel-check instructions.

This ensures both automatic enforcement and developer flexibility.

What’s Next: Supporting More Hardware, More Challenges

Our new runtime unlocks a new tier of algorithm complexity and speed — but we’re not stopping here.

Coming soon:

• Enhanced developer tooling: We’re working on better development tools to help authors optimize within the TIG runtime constraints.
• Expanded challenges: With native performance and multithreading, we can onboard a wider class of computational problems as challenges.
• Support for FPGAs and ASICs: We aim to make TIG the home for hardware-accelerated algorithm development.

Conclusion

Moving from WASM to native code execution has radically expanded what’s possible on the TIG platform. By combining the speed and flexibility of native code with lightweight instrumentation for fuel and signature checks, we’ve preserved verifiability with minimal compromise on performance.

This transition enables TIG to support the next wave of AI, optimization, and high-performance computing challenges — securely, transparently, and at scale.

ZUG, SWITZERLAND — 09 June 2025 — The Innovation Game is proud to announce a landmark evolution in our ecosystem: the formation of a Swiss Association established to accelerate our mission and redefine the lifecycle of innovation.

This is not just a legal step; it is the creation of a powerful engine for progress. The Swiss Association will be the cornerstone of our ecosystem, dedicated to onboarding, safeguarding, and commercializing high-value intellectual property from the world’s brightest minds. By establishing this entity in Switzerland, a globally respected bastion of neutrality and legal clarity, we are building a trusted bridge between pioneering academics and the commercial world.

For the TIG ecosystem, this marks the beginning of a new chapter. The Association is structured to attract and hold world-class scientific IP. Universities, research institutions, and top academics require a trusted, stable and reputable partner to house their discoveries, and the Swiss legal framework provides that assurance. This will create a powerful flywheel effect: the more we can attract and engage the best minds, the more we can capture and bring together the best ideas and intellectual property to provide a springboard for continuing innovation in our network. All of this adds up to creating potentially exponential innovative output from our ecosystem.

“We are moving beyond the theoretical. It’s time to build the infrastructure for a new, decentralized innovation economy,” said Dr. John Fletcher, PhD, a founder of The Innovation Game and Director of the newly established Swiss Association. “The formation of a Swiss Association to house The Innovation Game signals our steadfast commitment to this future. It will be a beacon for brilliant researchers, providing them with the resources to pursue breakthroughs within an open meritocracy. Crucially, it will then connect these breakthroughs with the market by licensing this IP to commercial entities, creating a self-sustaining cycle of value that will power the entire Innovation Game ecosystem.”

The Swiss Association has a clear and powerful mandate:

• Become a Global Hub for Intellectual Property: Actively attract and securely hold significant scientific and technological IP within the neutral and respected jurisdiction of a Swiss Association.
• Fuel Breakthroughs with Ecosystem Grants: Deploy treasury resources to provide substantial grants to academics, researchers, and challenge owners, funding the next wave of disruptive innovation.
• Drive Commercial Adoption through Licensing: Systematically license the Foundation’s portfolio of IP to commercial companies, turning groundbreaking research into real-world products and revenue streams.
• Guarantee Ecosystem Integrity: Oversee the continued development of our smart contracts, manage the community treasury to foster growth, and ensure the robustness of the proof-of-work validation that underpins our network.

The establishment of the Swiss Association is a critical step on our path to building a globally significant repository of intellectual property. We are creating a system where the world’s most important ideas can be nurtured, protected, and deployed for maximum impact, with the TIG community at its very core. The future of innovation is being built here, today.

As part of ongoing work to scale the TIG network and align token distribution, we are updating the token emission mechanism. The new mechanism was designed by TIG Labs in partnership with CryptoEconLab (CEL), recognised experts in token economics. This document, presents the first of two planned refinements to TIG’s tokenomics system. It focuses on adjusting token issuance to match the network’s current usage and growth, ensuring rewards scale with the number of active Challenges. The high-emission phase has delivered significant benefits including:

1. Token Distribution: Significant distribution of tokens via proof-of-work and innovation rewards has supported effective governance and robust participation in network decisions.
2. Validation of the OPOW Mechanism: A high token price during this phase stress-tested the novel optimisable proof-of-work (OPOW) mechanism, confirming its effectiveness under strong economic pressure.

Practical experience over the past 12 months has clarified the appropriate rate for adding Challenges. In response, we have developed a refinement to TIG’s incentive mechanism. This refinement is presented below. In the initial phase, effective token emissions are limited while Challenge numbers are low, with this cap lifted as the network scales. In essence, until the network scaled (which we define as the network featuring 100 Challenges) only a portion of new tokens will be distributed as rewards, with the rest held in a vault for future release after scale has been achieved (the number of Challenges exceeds 100).

1 Limitations of Token Emission Relative to Challenge Availability

Total value that can be injected into TIG ultimately depends on the number of Challenges, and TIG’s current incentive structure restricts the total value that can be injected into the protocol because the number of active Challenges is low. Excess token emissions when few Challenges exist create market pressure that lowers the incentive per token. This relationship is summarised by the flow:

Unconditional Supply Growth → Market Pressure → Reduced Incentive Power

In other words, rapid increase the token supply before the number of Challenges is scaled, results in reduced incentive for innovation, since market pressure tends to suppress token price. In order to address this, we are introducing a modification to the token rewards mechanism. The new mechanism (described below) ties token distribution to the number of active Challenges and diverts any unallocated tokens into a vault until the network reaches a set Challenge threshold. The expected trajectory of Challenge growth is also outlined, and a new “Challenge Owner” role is introduced to accelerate Challenge addition as the network scales. Finally, based on these growth projections, the document proposes a schedule for releasing the vault’s locked tokens, showing that the vault begins to empty, distributing additional token incentives, once the network exceeds the 100-Challenge mark.

2 Quantitative Overview of Changes

The token cap remains unchanged at 131.04M TIG, as does underlying emissions curve M(t) between 0 and 15.5 years: see Figure 3. While the number of Challenges n is below the threshold n = 100, a fraction Γ(n) ∈ [0, 1] of the emissions is diverted to a vault. Tokens in the vault will be distributed only after n > 100 (Figure 1). Figure 2 shows the expected number of Challenges that are added to TIG as a function of time (one round in TIG lasts on week); combining the rate of Challenge addition with the function Γ(n) allows us to plot the distribution of rewards in comparison to the underlying emissions curve (which remains unchanged): see Figure 3.

Figure 1: Ratio of minted TIG tokens distributed to the community (Γ) as a function of the number of Challenges. The distribution rate begins at Γ = 0.25 when there are 4 Challenges and increases as the protocol grows, reaching Γ = 0.5 at 15 Challenges — half the original distribution rate. At 100 Challenges, Γ = 1, matching the original schedule. Beyond this threshold, Γ asymptotically approaches 1.02, indicating the onset of vault depletion.

Figure 2: Projected growth of the number of Challenges on the TIG protocol over time. The number of Challenges increases quadratically, capturing the effects of network-driven growth. Projections are based on the introduction of the “Owner” role (see Section 3). Challenges arrive according to a Poisson process, the curve shows the median across 1,000 Monte Carlo simulations, with shaded bands representing the 5th to 95th percentile range.

Figure 3: Cumulative TIG tokens (in millions) distributed to the community under two emission schedules. The blue curve shows the original minting schedule, where tokens were immediately distributed, reaching half of the total supply by round 130. The green curve shows the new emission schedule, in which a portion of tokens are initially locked in a vault and distributed more gradually — reaching the halfway point by round 231. By round 800, all vault-held tokens have been distributed. The green curve represents the median across 1,000 Monte Carlo simulations, with shaded regions indicating the 5th to 95th percentile range.

3 Challenge Owners and Scaling of Challenge Addition

We are introducing a new role: Challenge Owners. Their responsibilities will include:

• Proposing and coding new Challenge instances with appropriate parameterisation.
• Engaging with the community on the TIG forum to refine and validate proposals.
• Updating Challenges to ensure the innovation remains aligned with real-world applications and competitive with state-of-the-art methods.
• Providing quarterly reports on performance against standard benchmarks.
• Recruiting expert advisors in relevant fields.

This initiative is designed to greatly scale the rate of Challenge addition, ensuring that the full token emission rate can be restored as network activity increases. The requirement to recruit expert advisors will also scale our involvement of domain experts, accelerating awareness of TIG within key technical communities.

4 Future Tokenomics Adjustments

The current change is the first of two planned refinements. The second change will address the allocation between standard Innovator rewards, breakthrough Innovator rewards, and Challenge Owners. Details of the second update are expected to be available in the next 6 to 8 weeks. These updates are aimed at maximizing efficiency of resource distribution, and aligning token incentives with tangible network growth.

In summary, tying token distribution to Challenge growth ensures TIG token rewards remain effective and fair as the network scales. This change fortifies the TIG economy for long-term success

During my PhD studies, I witnessed first hand how computation has become the backbone of modern research. Science has entered what is often called the 4th Paradigm — a new era of data-driven discovery powered by computation. My research, like that of many of my colleagues, relied heavily on computation — whether for simulating complex systems, processing large datasets, or automating experimental processes, computation is no longer just an advantage; it’s a necessity. However, this reliance on computation also means that scientific discovery is increasingly dependent upon — and therefore constrained by — the computational resources available.

Computation Catalysing Humanity’s Grandest Ambitions. From my research, I learned that it is computation that will bring us the types of technologies that will change the world. For example:

• Near-Limitless Clean Energy: Fusion research relies on real-time plasma control and vast simulations. A single breakthrough in computational capabilities could increase reliability and reduce costs by orders of magnitude, accelerating commercially viable fusion power.
• Interplanetary Colonisation: Advanced propulsion systems — nuclear, plasma-based — will require ultra-fast physics simulations to achieve real-time control.
• Cures for Diseases and Longer Lifespans: Drug discovery, genomics, and biological simulations depend on large-scale computational resources.
• Solutions to Climate Crises: Optimising energy grids, designing advanced batteries, and modelling Earth’s climate with unprecedented fidelity require computational breakthroughs.

It was clear that computation will be the key driver of progress in the 21st century, and to understand the trajectory of science and technology, I needed to understand the rate of progress in computational resources.

Hardware vs Algorithms: Two Titans of Compute

I began examining the main components that contribute to computational power: hardware and algorithmic methods. Hardware improvements, driven by companies like Nvidia, are incremental and relatively predictable, guided by well-established trends like Moore’s Law. The semiconductor industry attracts significant investment from the market, which suggests that it’s likely highly optimised. In contrast, progress in algorithmic methods is far more sporadic and unpredictable, often stemming from the insights of a small number of highly skilled researchers. This development process is largely dependent on government funding, which is relatively scarce and tends to be allocated inefficiently compared to private funding.

This seemed counterintuitive, because algorithmic advances can lead to improvements in computational capabilities that are orders of magnitude greater than those from hardware alone. Consider, for instance, the attention mechanism that sparked the current AI revolution, or the Fast Fourier Transform; every truly game-changing leap in computation has its origin in an algorithmic breakthrough. Yet, despite their immense potential, algorithms receive far less publicity and private investment than hardware. This disparity struck me as odd. If algorithms hold the key to unlocking massive computational gains, why is their development so underfunded and erratic? The answer, I discovered, lies in the economics of algorithm development.

If Algorithms Are So Important, Why Haven’t I Heard About This Before?

The unpredictability of breakthroughs makes algorithmic research a high-risk activity, deterring private investment. Historically, this has left the field reliant on government funding, which, while promoting openness, is underfunded and inefficiently allocated. However, I was surprised to learn that economists of science had long warned against disrupting this state of affairs, despite the significant drawbacks of reliance on public funding. The reason for this is fascinating: for the most fundamental areas of science, the risk posed by monopolies is considered to be so serious that it outweighs the potential benefits that increased funding and market-driven efficiency can bring. For example, if algorithmic research were to attract private funding, developers would likely keep their breakthroughs proprietary to maximise returns. Given the power of algorithms, this could lead to monopolies, where a single entity gains an unassailable advantage, ultimately killing competition and innovation.

This is why the public rarely hears about algorithms. Instead, it’s the multi-hundred billion dollar investments in hardware that make the headlines. The risk inherent in algorithm development means that commercial ventures tend to leave this side of things to the public sector and differentiate with respect to other factors, notably hardware. It follows that companies naturally want to publicise investments like hardware at every opportunity, since this will tend to signal that they are ahead of, or at least keeping up with, their competition. Thus, the public’s perception that hardware is more significant than algorithms has developed not because algorithms are less important — indeed, you hear less about them precisely because they are recognised as too crucial to risk privatising! Given my understanding of the immense power of algorithms, this rang true. I had finally found the explanation I sought. But my uncertainty was replaced with another question: what about AI?

The Rise of AI Agents

It has long been assumed that AI will eventually be capable of doing most tasks that humans can. The development of algorithms is no different. It was clear that algorithmic research would be transformed; no longer constrained by the limited pool of capable people, it would become far more predictable and scalable, and ultimately viable for private investment. Clearly, in this scenario, the traditional policy of inaction intended to reduce the risk of private algorithms would no longer be tenable. However, without a suitable system of incentives in place, this seemed a dark fate indeed: where powerful algorithms are privatised, one entity could eventually gain an accelerating advantage, centralising entire industries and even science itself. It was this realisation that compelled me to devise such a system. I called it The Innovation Game.

The Innovation Game: A Solution

The Innovation Game (TIG) is a network with a single objective: to accelerate science and technology by incentivising the development of open algorithms. TIG creates a market for open algorithms, encouraging innovation while ensuring that, whether developed by human researchers or AI, the results remain available to all. Drawing inspiration from the open-source software movement and Bitcoin, TIG establishes a system where innovators are rewarded for their contributions through a token-based economy, but only if their algorithms are adopted and used widely. This mechanism means that the most effective algorithms rise to the top, while openness prevents any single entity from monopolising the field. TIG’s hybrid licensing model means that algorithms remain open for anyone to examine and iterate upon, while enabling value capture from commercial applications.

Openness and Decentralisation in TIG

As a researcher, I’ve always intuitively valued the openness of scientific knowledge. But with TIG, openness is much more than a personal preference; it is integral to the mechanism by which TIG seeks to accelerate science and technology. The fact that all algorithms on TIG are open ensures:

• Fast, Cumulative Innovation: Open algorithms let the community build on each advance, stacking improvements rather than starting anew. This speeds up discovery in ways closed systems can’t match.
• Global Coordination: Like Bitcoin mining, TIG is open to all — researchers, AI agents, anyone — without gatekeepers. This spreads participation wide, keeping progress diverse and resilient.
• Monopoly Resistance: Similar to how Bitcoin uses vast computational power to defy centralisation, TIG taps the collective intelligence of its network to make monopolising algorithms impractical. No single player can dominate when the system thrives on scale and openness.

It is important to note that, although TIG anticipates the AI era in science, the benefits it brings are completely independent of whether algorithms are developed by people or AI. For researchers, TIG could be viewed as a sort of perpetual hackathon, to which they can submit improvements and earn rewards. AI agents, on the other hand, can “mine” algorithms, submitting any promising candidates to TIG. In all cases, algorithmic development benefits from the efficiencies of market allocation and the capital that comes with private investment, while all results remain open forever. Without TIG, these breakthroughs would likely stay private, leading to the monopolies economists dread. With it, we can finally achieve the combination of factors that has eluded science for centuries: value capture and shared knowledge.

TIG: The Future Is Open

By radically lowering barriers to algorithmic progress, TIG reignites revolutionary science in the spirit that propelled humanity from steam engines to spaceflight. By making algorithms investable while keeping them open, TIG addresses the dual challenges of funding inefficiency and monopoly risk. It provides a sustainable, market-driven framework for accelerating scientific progress, ensuring that the computational tools we rely on continue to evolve rapidly.

In an era where computation is the key to unlocking humanity’s greatest challenges — from clean energy to interplanetary travel — TIG offers a way to reward researchers as if they were entrepreneurs while maintaining the openness essential for collective scientific progress. It’s a vision that resonates deeply with me as a scientist, and I believe it will shape the future of research for generations to come. The vision of The Innovation Game is already attracting the world’s top experts in computational science, technology, and intellectual property to the TIG network. We invite you to join us in this project to shape the destiny of humanity and set science free.

Over the last few months, a team of giga-brains have been assembled. Based in Cambridge UK, this group forms the beginnings of TIG Labs, the intellectual engine that will drive The Innovation Game.

These individuals have proven academic merit of the highest order, and have been collected from the World’s leading institutions. They are the beginning. There will be more.

In the following article, we introduce you to TIG labs- a team that are working tirelessly to advance the protocol’s mission of accelerating scientific and technological breakthroughs, through drastically speeding up the rate of algorithmic innovation.

The mission

The goal of TIG Labs is clear: bridge the gap between theoretical computer science and practical problem-solving. In short, working with academics and industry to take computational problems of the highest significance, and convert them into TIG challenges that will incentivise finding solutions that can be applied in the real world to great effect.

Understanding Challenges

To understand quite how important a role TIG Labs will be playing, we need to understand what TIG challenges are.

A Challenge in the TIG protocol represents a well-defined computational problem- where finding a solution requires significantly more computational effort than verifying one. TIG focuses on challenges for which better solutions will have the largest positive impact on the human experience.

Examples of such challenges include optimisation problems in machine learning, logistics, medical imaging, and computational drug design.

While TIG Labs currently designs and implements initial challenges, the protocol will evolve to enable anyone to create a challenge through a decentralised framework- this is outside of the scope of this article, but there’s more on that to come!

These Challenges serve as the foundation for innovation within the protocol, allowing participants to contribute novel algorithmic solutions that can be objectively measured and rewarded.

Current Challenges can be explored at https://www.tig.foundation/challenges.

The Team

TIG Labs has assembled a distinguished team of researchers representing some of the world’s most prestigious academic institutions:

• Dr. John Fletcher (CEO, Chief Scientist and Co-founder) — If you know TIG, you’ll know The Doctor: A University of Cambridge PhD graduate in mathematics and theoretical physics, Dr. Fletcher holds more than 30 patents in cryptography and distributed systems. His pioneering work in distributed incentives spans nearly a decade, establishing him as a leading thinker in the field.
• Kilian Scheutwinkel (Senior Researcher & Computational Scientist) — A University of Cambridge Astrophysics PhD candidate, Kilian brings valuable industry expertise through his work with Nethermind’s security research team and the venture capital firm TRGC.
• Aoibheann Murray (Computational Scientist) — Aoibheann graduated from Trinity College Dublin with a First Class Honors in Theoretical Physics. Following this, she secured a Masters from the University of Oxford’s Mathematical Modelling and Scientific Computing program, finishing top of her class.
• Myron Sukhanov (Computational Scientist) — A Physics Olympiad medalist conducting research at the Cavendish Laboratory, University of Cambridge — a department famous for discoveries you may have heard of- including, but not limited to, the electron, the neutron, and the structure of DNA.
• Jack Fan (Research Strategist) — Currently pursuing an MSci in Physics with Theoretical Physics at Imperial College London, an institution consistently ranked among the world’s best for physics and engineering.
• David Liu (Computational Neuroscientist) — A Cambridge PhD researcher at the Computational and Biological Learning Lab, with an impressive academic record including a Quadruple First Class from Cambridge’s Natural Sciences program.

Impact on The Innovation Game

The team’s collective experience positions TIG Labs as a centre of excellence in computational innovation. Their work is vital to the success of The Innovation Game in several key ways:

Research Leadership

Cambridge is a global hub for computational research- TIG Labs is leveraging its proximity to world-class research facilities and academic networks to:

• Develop cutting-edge Challenge frameworks
• Advance theoretical understanding of asymmetric problems
• Foster collaborations with leading academic institutions

Bridge to Academia

A key part of the TIG Labs role is liaising with academics, and bridging their expertise with real world, applied challenges. As such, the team will procure:

• Direct access to emerging research trends
• Collaboration with leading research groups
• Attraction of top-tier talent to the protocol
• Validation of challenges by respected academic experts

Current Focus

Finally the current focus- TIG Labs is currently advancing research in several frontier areas, including:

• Zero Knowledge Proof (ZKP) generation optimization
• Advanced AI optimizer development
• Hypergraph partitioning challenges
• We’re excited to share some of the work that has been done here, and will be doing in short order!

Looking Forward

This is a scientific revolution.

For the first time, we have a protocol that can actively co-ordinate innovators and compute providers worldwide, directly focusing their collective efforts on advancing Science and Technology.

This isn’t where we’re stopping.

TIG Labs facilitates an additional level of co-ordination- connecting the distributed power of the protocol with academia and industry, to ensure that the protocol challenges align with real scientific needs. This means that the most impactful problems will be identified, and that the output of the protocol is tangible scientific and technological progress.

The future is bright.

This piece introduces the concept of “Breakthrough Rewards”, which will ensure that Innovators who create Breakthroughs may be substantially rewarded.

Method vs Implementation

Algorithmic methods (referred to as “Methods”) can be thought of as the description of an approach to solving a computational problem.

Algorithmic implementations (referred to as “Implementations”), takes a Method and implements it as executable code.

TIG considers Methods to be Breakthroughs if they meet certain criteria, outlined in the following sections.

Breakthrough Rewards

Breakthrough rewards are available to incentivise the creation of Breakthroughs.

Breakthroughs should have the following attributes: i) Patentability, and ii) Performance Improvement.

Patentability requires that the algorithm a) be ‘novel’- i.e completely new, and b) be ‘inventive’- i.e. non-obvious and capturing some ingenuity. These properties will be determined by token-weighted vote.

Performance Improvement is self explanatory, and will be indicated by Benchmarker adoption.

The New Reward Structure

Breakthrough Rewards will be an additional tier of rewards which constitute 15% of total emissions. These will be available to Innovators who submit Methods that get voted as eligible for Breakthrough Rewards.

Previously, emissions were split between Benchmarkers (85%) and Innovators (15%), with no distinction between whether an innovation was a Breakthrough, or an Implementation.

Now, 70% will go to Benchmarkers, 15% to incentivize Breakthroughs, and 15% to Implementations.

Why This Change Matters

Breakthroughs are difficult. They require a sustained period of work and/or flashes of genius. The protocol must ensure that contributors are maximally incentivised to submit their Breakthroughs if they have them.

• In TIG 1.0, Innovators could only submit Implementations, even if that Implementation embodies a Breakthrough. A minor code optimisation to these Implementations could result in the original Innovator losing their rewards to other code optimisers, despite the underlying Breakthrough’s greater significance.
• This served as a disincentive for innovators to put a lot of time and effort into coming up with Breakthroughs. A number of community members highlighted this!
• Going forward, Innovators can directly submit Breakthroughs to TIG and receive a continuous flow of rewards, separate from any Implementations of that Breakthrough. These rewards continue until another Innovator submits a better Breakthrough for the same Challenge, provided that it passes the criteria.

With this change, Innovators who invent Breakthroughs will earn rewards for a much longer period than previously possible.

Breakthrough Rewards v Implementation Rewards

To reiterate, Breakthrough Rewards will now be considered as distinct from Implementation Rewards. Breakthrough Rewards will relate to a truly innovative/novel method for solving a Challenge.

Implementations will continue to be rewarded in the same way. Any implementation that gains 25%+ of adoption or has been merged, will earn rewards.

How to Get Breakthrough Rewards

Simply put, an Innovator submitting a would-be Breakthrough must submit a claim with supporting evidence that they have made a Breakthrough. This will then be voted on by the community!

Below is a walkthrough of the process from an Innovator’s POV:

STEP 1: The Innovator must identify a Method they believe to be a Breakthrough, before submitting the Method together with supporting evidence that it is a Breakthrough.

STEP 2: A fee must be paid to make the request- this ensures that only serious requests are made.

Note: Any Innovator can then attribute new Implementations to this Method. It is highly recommended that an Innovator submitting a Method also separately submits an Implementation attributed to that Method (so voters can more easily assess the performance improvement criteria for a Breakthrough).

STEP 3: During the Round in which the request was submitted and the subsequent Round, the Method and supporting evidence will be kept confidential (“Escrow Period”).

STEP 4: After the Escrow Period, the Method and supporting evidence will be made public to allow for the community to scrutinise it (“Scrutiny Period”).

STEP 5: During the Scrutiny Period, $TIG token holders may vote on whether the Method is eligible for Breakthrough Rewards. The number of votes is based on the token holder’s deposit, and vote tallies are finalised at the end of the Scrutiny Period.

STEP 6: If a Method is voted as eligible for Breakthrough Rewards, the Method can be “adopted” (summing adoption of all Implementations attributed to that Method). A Method must reach 50%+ adoption (or be merged) in order to start earning Breakthrough rewards. Similar to Implementations, Methods exceeding the adoption threshold will also earn a merge point every block, and may be merged if it earns sufficient merge points over a Round. If a Method is merged, it is considered a Breakthrough.

What If There Are No Breakthrough Rewards

For any Challenge, the Breakthrough Rewards might not be fully allocated to Methods. This can occur if there are Implementations that are attributed to Methods/Breakthroughs outside of TIG. Any unallocated Breakthrough Rewards will go into a ‘bootstrap fund.’ This fund will be used exclusively to bootstrap the development and submission of Breakthroughs to TIG, especially for Challenges where state-of-the-art Methods/Breakthroughs exist outside the protocol. For example, it could fund world-leading experts to research and develop more efficient Methods for specific TIG Challenges.

Conclusion

The introduction of Breakthrough Rewards underlines the protocol’s commitment to providing a real incentive to anyone in the world that is capable of contributing a Breakthrough, to do so.

This illustrates one of TIG’s cornerstone objectives: to incentivise the best talent to contribute innovation to TIG and make the protocol the default source for the best algorithmic methods.

It also reflects the TIG team’s commitment to listening to the community and taking action to actively evolve the protocol as a result.

TIG will become the de facto hub for algorithmic innovation and we’re confident that this is a big step forwards towards achieving that aim.

Learn More:

For an additional resource on Breakthrough Rewards, you can watch our Community Call recording on YouTube which addresses key questions brought up by community members.

Proof of Deposit represents a *huge* leap forward for the TIG protocol.

At first glance, it’s a system of economic incentives that drives more responsible mining practices.

A deeper look reveals an intricate mechanism that serves to ensure that benchmarkers are stalwart defenders of the protocol, and that TIG token holders have real power to make a material impact for a material reward.

Perhaps most intriguingly, it creates a new layer to the game, where benchmarkers must balance compute, reputation and community support to succeed.

In the following article, we begin with an overview of the key points of the update, before moving into examining it in greater depth.

Overview

A) Benchmarkers must now have skin in the game, and as such be more aligned with the long-term interests of the protocol.

A new benchmarker must lock up 100 TIG tokens to begin being rewarded for their qualifying solutions (qualifying solutions being ones eligible for reward). Every 100 TIG they lock raises the cap on the number of qualifying solutions that can count by 1. At 10,000 self-delegated TIG, they have the ability to take external delegation.

B) TIG token holders can now have a material stake in the protocol, through delegated deposit.

TIG token holders can now delegate TIG tokens to a benchmarker. This involves choosing an amount of TIG tokens to lock, and then choosing an amount of time to lock them for. A weighting score will be applied based on the number of weeks of locking period remaining at any given time- this will constitute ‘weighted deposit.’

Benchmarkers will choose which fraction of their rewards to share with delegators, which will be distributed pro-rata each block according to which delegators contributed the most to ‘influence.’

The bigger the weighted deposit , the more contribution to benchmarker influence will have been made.

C) Benchmarking compute and deposit must be proportional to avoid imbalance penalties.

Benchmarker rewards are based on ‘influence.’ Influence is calculated from the fraction of qualifying solutions computed by a benchmarker, together with the fraction of total weighted deposit that has been delegated to their account.

In short, this means that Benchmarkers’ share of the compute of the network must be balanced with their associated deposit. If they want to maximise their earnings from increased compute, they must increase deposit.

D) Benchmarking is even more strategic.

Influence is a constant-sum game. If one benchmarker’s influence increases, another’s decreases. This creates a strategic ecosystem where benchmarkers must balance their compute power, deposit levels, and delegator relationships to maximize rewards.

PoD in greater depth

The core idea: skin in the game

At its core, Proof-of-Deposit ensures that Benchmarkers (miners) have some skin in the game. If Benchmarkers want to find and submit solutions to challenge instances, they will need to lock up tokens in order for their compute to count.

Think of it as putting down a security deposit before renting an apartment — it ensures everyone plays by the rules!

How Does It Work?

The two core concepts here are self-delegation, and external-delegation.

Self-delegation

If you want to become a miner in TIG, you will need to do the following:

1. You’ll need to lock up some TIG tokens (at least 100)

2. The tokens must be locked for at least a week.

3. For every 100 TIG tokens you lock up, one more qualifying solution per challenge qualifies for rewards (every 100 tokens raises the cutoff cap by 1).

To illustrate- if a benchmarker submits 10 solutions which would otherwise qualify, and has only delegated 100 TIG, only one of the solutions will qualify. They will have 10 qualifying solutions if they have 1000 TIG staked.

External-Delegation

Here’s where it gets interesting, and the games begin.

If benchmarkers self-delegate 10,000 TIG, they unlock the ability to receive delegation from others to support their mining operation.

Token-holders will be able to lock up TIG token as a ‘deposit.’ They will then be able to delegate these tokens to Benchmarkers and receive a fraction of their mining rewards.

Deposits aren’t locked up for the entire time a delegator specifies. They are vested linearly over the time period, meaning that delegators can claim a fraction of their deposit back each week.

Benchmarkers choose how much delegators get- ranging from 0–25% of total rewards. This fraction can be changed once every 1440 blocks.

Rewards will be distributed pro-rata based on each delegator’s weighted deposit (calculated from the number of tokens delegated, and the amount of time those tokens have been locked for- specifically how much of their locking time is remaining).

Delegators will be able to cancel or re-delegate their tokens if they feel that a benchmarker isn’t performing up to scratch. They will be able to do this once every 60 blocks at most!

Understanding Deposit Weighting

Weighting is an absolutely core component to the notion of deposit.

Weighting is based on the number of weeks remaining on a deposit’s locking period.

The influence calculation, which determines how big a share of the mining rewards that a benchmarker receives (more on that below), relies on ‘weighted deposit’- not merely the amount of tokens deposited, but the amount of tokens deposited combined with how long they have been locked for- specifically, how much of their locking period is remaining.

In short, externally-delegated, and self-delegated tokens will count more towards influence (detailed below) the longer the remaining locking time is.

How weighting is calculated

Deposit weighting is calculated based on how many weeks your tokens have remaining on the locking period.

Maximum weighting score will be assigned to the fraction of your tokens which have a remaining locking period of at least 26 weeks.

If you lock up 100 tokens for 52 weeks, you’ll receive a maximum weighting score on the proportion of your tokens that have 26+ weeks of lock left, which would be approximately half.

If you lock up 100 tokens for 26 weeks; maximum weighting score would only apply to 1/26 of your tokens- as only that fraction would have 26+ weeks of lock left.

This means that locking 100 tokens up for 52 weeks would give a higher weighting score than locking them up for 26.

Let’s look at some examples:

A token with a remaining locking period of 1 week would have a weighting score of 1.

A token with a remaining locking period of 26 weeks would have a weighting score of 26.

Let’s compare locking 100 TIG for 1 week, 6 months, and one year.

This lock would be a linear vest, which would start from the point of locking, and end at 1 week, 6 months, and one year respectively.

The respective weighting scores would be 100, 1350 and 1975.

To be clear: as the remaining lock-up time decreases, so does PoD weighting!

Why deposit matters: Influence and rewards

Deposit matters because it is now a fundamental part of the calculation of ‘influence.’

Influence determines how big a slice of the weekly mining rewards that a Benchmarker will get.

Influence is calculated from the amount of compute a benchmarker has (as shown by the fraction of total qualifiers that they have submitted across the challenges), together with the fraction of total weighted deposit delegated to their mining account.

The balance requirement

Influence is penalised when there is an imbalance.

Benchmarkers receive an ‘imbalance penalty’ if the fraction of total qualifying solutions that they’re responsible for (as part of the total number of qualifying solutions submitted by benchmarkers) is disproportionate to the fraction of weighted deposit that has been delegated to their mining account.

Crucially relevant to PoD, there will be imbalance when the Benchmarker’s share of compute is not proportional to their share of total weighted deposit. There *must* be deposit and compute alignment, with the fraction of total compute that a particular miner has, being proportional to the fraction of total deposit associated with their account, with a maximum ratio of 1:1.25.

As an example: If you’re contributing 10% of the network’s computing power, you should aim to have about 10% of the total weighted (up to a maximum of 12.5%).

In summary, POD introduces an interesting balancing act. Your mining power should match your deposited tokens. If you’re mining too much with too little deposit you’ll face imbalance penalties.

Competitive dynamics

This update creates an intricate competition between benchmarkers for network influence. As a reminder, influence determines which share of the fixed reward pool (currently 42.5 TIG per block until the next halving), that a benchmarker receives.

Influence is a constant-sum game, as the sum of influence across all benchmarkers equals 1.

This means that when a benchmark increases their delegated deposit, it automatically reduces other benchmarkers’ fraction of total deposit, and thus reduces their influence.

To maintain their reward share, benchmarkers must either:

1. Increase their own deposit through additional self-delegation or,
2. Attract more delegators.

A benchmarker could conceivably attract new delegators in a variety of ways, most obviously through a) offering more competitive reward shares, b) building a reputation as a trustworthy operator and making delegators aware of this, or c) attracting new TIG token holders who are confident in their reputation.

Conclusion

PoD announces the arrival of TIG 2.0. The protocol has been constantly evolving, and shall continue to do so until it is *the* defacto powerhouse of algorithmic innovation- driving critical progress across Science and Technology.

This is a new sort of protocol- one which acts intentionally in co-ordinating the world and directing its output towards achieving critical outcomes.

We’re lifting the limit on what’s possible, and this is just the beginning.

Join us, and help make it happen: https://discord.gg/agkDkG8v62