Proposal #113: Distribute v0.2.16 bounty rewards

Bounty Payments Associated with v0.2.16 This proposal covers 104,150 USDT from community funds for development, research, infrastructure, security reporting, and code review work on the Gonka protocol. The work includes contributions to upgrade v0.2.16 and larger projects spanning multiple upgrade cycles. 1. Decode-PoC: 38,000 USDT. Decode-PoC moves from prefill-only proof of compute covering token generation, aiming to make measured compute capacity better align with real inference workloads. Axel-t published the [original proposal](https://github.com/gonka-ai/gonka/issues/1135) in April 2026. The work covers CUDA graph support, vLLM porting, and testing across hardware configurations and models. Axel-t: 18,000 USDT for CUDA graph support. Technical details are available in the [PoC-decode documentation](https://github.com/axeltec-software/vllm/blob/poc-v0.20-decode-poc-cg/benchmarks/poc/POC DECODE.md). - Previously paid: 30,000 USDT on June 17, 2026, through governance proposal 76, for the scheme's design and experimental feasibility validation. - Upon completion, a final bounty of 12,000 USDT for Axel-t's contribution to the vLLM port, test suite, and on-chain integration will be proposed together with the upgrade that integrates decode-PoC on-chain. Both the previous payment and the planned final bounty are outside this proposal's total. Kaitaku.ai: 20,000 USDT for porting, experiments, validation, and integration work. The [initial report](https://github.com/kaitakuai/experiments/blob/main/reports/2026-06-decode-poc-research-and-integration.md) covers June 9-September 10, 2026, with [continued integration work](https://github.com/kaitakuai/experiments/blob/main/reports/2026-09-decode-poc-glm-028-and-0300-migration.md) documented through September 25, 2026. 2. gonka-poc plugin: 5,000 USDT. This project packages PoC functionality as a vLLM plugin, reducing the work needed to support new vLLM releases. Kaitaku.ai: 5,000 USDT for plugin development and integration. The [report](https://github.com/kaitakuai/experiments/blob/main/reports/2026-07-gonka-poc-plugin-and-residual.md) covers June 16-September 8, 2026. 3. Model benchmarking and integration: 13,000 USDT. This work prepares candidate models for use on Gonka. It includes necessary vLLM updates, PoC benchmarks across common GPU types, and testing validation thresholds to detect model substitution while allowing different GPU types to validate each other. It also includes on-chain proposals with model parameters and coefficients. Kaitaku.ai: - DeepSeek-V4-Flash: 5,000 USDT for experiments, a vLLM update, an on-chain model proposal, and additional experiments for the 0731 version. - GLM-5.3-Flash: 4,000 USDT for experiments, a vLLM update, and an on-chain model proposal. - GLM-5.2: 2,000 USDT for experiments and deployment images for B300, B200, and H200. - Hy3: 2,000 USDT for experiments. This initial evaluation did not result in an on-chain proposal. The proposed amounts use the following breakdown: 2,000 for a set of experiments; 1,000 for a vLLM update associated with a new model proposal; and 1,000 for preparing the on-chain proposal with model parameters and coefficients. DeepSeek includes an additional 1,000 for repeating experiments for the 0731 version, without any additional updates specific for that version. 4. Vulnerability reporting: 5,000 USDT. This payment recognizes the report of a vulnerability that could halt the chain. @vitaly-andr: 5,000 USDT for the high-severity vulnerability report [ 1205](https://github.com/gonka-ai/gonka/issues/1205), with a proposed fix in [cosmos-sdk 16](https://github.com/gonka-ai/cosmos-sdk/pull/16). 5. Trainshards: 7,000 USDT. This work contributes to the development of the network's training capabilities, including contributions [ 1350](https://github.com/gonka-ai/gonka/pull/1350) and [ 1618](https://github.com/gonka-ai/gonka/pull/1618). @x0152: 7,000 USDT for eight weeks of part-time Trainshards v0 development. 6. ML-node ob…