TokenWorks announced the $CMD token, which uses a randomized Frankenstein approach to build its smart contract rules. It is described as a “Frankenstein-esque smart contract” to democratize token creation in the era of vibecoding.
Users submit rule ideas e.g., no trading on weekends or add a 1% tax that buys and burns every 24h before launch, paying a fee to participate. Over 10 rounds, one idea per round is randomly selected on-chain and automatically applied to the smart contract using AI. This creates a quirky, unpredictable stitched-together rule set. Each winning idea gets 1% of the final token supply. 10% of the total supply is distributed among all participants.
Participation started on April 10, 2026. This seems like a fun, community-driven experiment in meme and token mechanics—letting the crowd and randomness + AI patch together unusual features like trading restrictions, taxes, burns, or other behaviors directly into the contract
AI rule integration in the $CMD token from TokenWorks is the mechanism that automatically turns community-submitted ideas into live, enforceable code inside the final smart contract. The project runs 10 rounds each lasting 12 hours. In every round: One user-submitted rule and idea is randomly selected on-chain using verifiable randomness, likely something like Chainlink VRF or a similar oracle.
Register for Tekedia Mini-MBA edition 20 (June 8 – Sept 5, 2026).
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register for Tekedia AI Lab.
That selected idea is then automatically translated and applied to the evolving smart contract using AI. After all 10 rounds, the token launches with the 10 stitched-together rules baked in — creating the quirky Frankenstein token. This is what makes it different from normal token launches: instead of developers manually coding every feature, AI acts as the surgeon that patches each winning idea into the contract.
Users pay a fee and submit natural-language ideas. These are stored on-chain or in a linked system. At the end of each round, a random winner is chosen transparently on the blockchain. This ensures fairness and prevents any central party from picking favorites. An AI model or a pipeline of models takes the winning natural-language description and converts it into Solidity or the relevant smart contract language code. The AI must understand the intent of the rule.
It generates the necessary functions, modifiers, hooks e.g., in _transfer, beforeTokenTransfer, or custom logic. It handles edge cases, security considerations, and integration with existing code. The generated code snippet is then automatically inserted into the base smart contract template. This could happen via: On-chain execution (if the contract is upgradeable or uses a modular/proxy pattern).
An off-chain service that compiles and verifies the new contract version and deploys and upgrades it with on-chain verification or multisig for safety. A system where the contract is designed to be highly modular from the start, and AI helps compose the modules. After all 10 patches, the complete contract is finalized, audited (presumably), and the token launches with the full set of Frankenstein rules active.
The end result is a token whose mechanics are a random collage of community ideas — some might synergize nicely, others might create weird or funny interactions. Non-technical users can propose sophisticated mechanics without knowing how to code Solidity. Speed: 10 rounds in ~5 days (10 × 12 hours) would be impossible if humans had to manually implement and test each rule.
AI can interpret vague or creative vibecoding ideas and turn them into functional code. Scalability: Handles the unpredictable nature of what wins each round. AI-generated smart contract code can contain bugs, vulnerabilities or unintended behaviors — especially when multiple rules interact in unexpected ways. This is a high-risk meme and experiment token. TokenWorks will likely have some verification, testing, or human oversight layer, but details aren’t fully public yet.
Patching a contract 10 times can increase complexity and deployment costs. The selection is on-chain for transparency, but the AI step itself may have some off-chain component, common in such hybrids. The exact technical implementation which AI model, how the patching is executed on-chain vs off-chain, whether the contract uses upgradeable proxies, etc. hasn’t been fully detailed in the initial announcements, but the core promise is clear: random community idea ? AI ? live rule in the contract.
Bittensor’s TAO Price Decline is Connected to Covenant AI Leaving the Network
Bittensor’s TAO token dropped sharply around 15-18% or more intraday, with some reports noting peaks near 27% in short windows after Covenant AI announced its exit from the network. Covenant AI; operating high-profile subnets like Templar/SN3 for large model pre-training, Basilica/SN39, and Grail/SN81 was one of Bittensor’s biggest and most successful contributors.
They gained attention for training the Covenant-72B model decentralized across nodes, which helped fuel a strong TAO rally, up ~90-100% in recent weeks/months and even drew praise from NVIDIA’s Jensen Huang.
In a public statement Convenant founder Sam Dare accused Bittensor of decentralization theatre. Governance power is concentrated with co-founder Jacob Steeves who allegedly controls the triumvirate and makes unilateral decisions. Specific grievances: suspension of emissions to their subnets removal of moderation privileges in community channels, unilateral deprecation of subnet infrastructure, and timed large TAO token sales that applied economic pressure during disputes.
They concluded they could no longer responsibly build or raise capital in the ecosystem and chose to exit rather than pass risks to supporters. The market reacted fast: TAO fell from around $330-340 to lows near $263-285 within hours, wiping out hundreds of millions in market cap; some estimates $900M in a sharp session and erasing much of the recent rally. Trading volume spiked significantly, with added pressure from liquidations and reports that Covenant-related wallets sold ~37,000 TAO. Subnet-specific tokens dropped even harder.
Price has since partially recovered in the volatile session hovering around $260-300 depending on exact timing, but sentiment took a clear hit. Bittensor is a decentralized AI network where subnets compete for emissions based on performance and staking. Subnets like Covenant’s demonstrated real technical wins in decentralized training, but the project has long faced questions about actual decentralization vs. founder influence, economic sustainability and governance.
This isn’t the first internal drama in crypto projects, and exits or disputes happen when incentives or control clash. Covenant’s success made their departure particularly visible and painful for confidence in the decentralized AI narrative. More volatility and FUD are likely.
TAO is testing supports; further downside risk exists if other subnet operators voice similar concerns or if liquidations cascade. Broader market factors didn’t help. Depends on how the core team responds, whether other builders stay or leave, and if Bittensor can show genuine progress toward distributed governance + real-world AI utility and revenue. The halving mechanics and institutional interest were prior tailwinds, but trust is now a bigger variable.
Covenant operated major subnets responsible for an estimated ~9% of network emissions. Their departure creates a short-term vacuum. Other teams may compete for the slots, but replicating specialized work (decentralized large-scale pre-training, compute scheduling, fine-tuning) won’t happen overnight. Assets like trained models and contributor networks could leave with them.
As one of Bittensor’s most visible and technically successful contributors; praised even externally, the exit raises questions about governance and incentives. It fuels debates on whether the network truly delivers decentralized AI or risks decentralization theatre. This could deter new builders or capital raises for projects staying in the ecosystem.
Critics point out that high emissions and subsidies have funded development, but external revenue remains low e.g., top subnets generating far less than the subsidies paid out. If builders can exit after benefiting from rewards, it challenges the value proposition for TAO holders and stakers.
This highlights a classic crypto tension: hype around decentralization can collide with practical control and incentives. Crypto moves fast, and one high-profile exit doesn’t necessarily doom a network, but it does force a reality check.