TIG: Token 2049 Recap

The Spaces recaps John’s trip to Token2049 Singapore, highlighting substantive conversations with Vitalik Buterin, Jacob Steeves (Bittensor), and Joe Lubin, and how Tig (The Innovation Game) is positioning algorithms as a public good. The discussion contrasts the investability and predictability of hardware with the overlooked but equally crucial role of algorithms, arguing that automated, industrial-scale algorithm discovery will transform innovation—and must be open to prevent dangerous centralization of power. John explains why Tig is not a Bittensor subnet, noting different anti‑sybil/security models (efficiently verifiable proof‑of‑work vs. stake), while emphasizing strong synergy and prospective collaboration (e.g., algorithm mining via agents like AlphaEvolve). The Optimizer Challenge—framed as one of Tig’s most important tasks—launches next week, targeting a core AI training primitive (optimizers) with major implications for open model training across ecosystems. Academics are increasingly engaging with Tig because the decentralized token mechanism is necessary to make open algorithm markets viable. The session closes with a raffle, a brief update on ResearchCoin (Brian Armstrong’s initiative), and openness to connect with Bittensor quantum subnet leaders.

Token 2049 Singapore recap; TIG–Bittensor synergy; algorithmic decentralization; and the Optimizer Challenge

Participants and roles

• John: Speaker 1; founder/lead of TIG (The Innovation Game). Primary technical voice in the discussion.
• Sparta (host): Speaker 2; VC-side perspective, moderates the Space, runs raffle, produces explainers/videos. Likely goes by the handle “Sparta.”
• Noted attendees mentioned: Dread Bongo, Adele, Magnate. Raffle winners: Magnate and “0”.
• External figures referenced: Vitalik Buterin (Ethereum), Jake (Jake Steves/Stevens) from Bittensor, Joe Lubin (Ethereum/ConsenSys), OpenTensor Foundation reps, Brian Armstrong’s ResearchCoin team, Dr. David Liu (Meta; Cambridge; optimizer research), Thibault (leading researcher, state-of-the-art vehicle routing), DeepSeek (Chinese open-weights model).

Token 2049 Singapore highlights and networking

• John’s objective and assessment:
  • Attended Token 2049; emphasized the high density of valuable side-events and private gatherings for networking.
  • Reiterated personal trade-off: travel is exhausting, but ROI of connections and in-depth conversations is meaningful.
• Key meetings:
  • Vitalik Buterin: Reconnected; updated Vitalik on TIG (conceived ~10 years ago). Discussed AI safety and open-source oriented movements. Vitalik is engaged in AI safety/open-source maximist initiatives.
  • Jake (Bittensor): Deep discussion of incentives and shared goals. Strong conceptual overlap in vision between TIG and Bittensor, with complementary technology layers.
  • Joe Lubin: Brief greeting; he was focused on treasury/financial events. John noted he should have spent more time, given Lubin’s stature.
  • Numerous other protocol teams: Many were hearing TIG’s approach for the first time; conversations were typically deep and sustained, reflecting genuine interest.

Why algorithms are as critical as hardware (and increasingly investable)

• Sparta’s framing: Current visible AI narratives (chips, data centers) dominate. Algorithms are less visible yet fundamental. He asks whether algorithms are arguably “the most fundamental component” of modern tech.
• John’s position:
  • Historically, algorithmic improvements have contributed at least as much speedup as hardware (often comparable or greater), but hardware’s predictability (Moore’s Law) made it more investable.
  • Algorithms are application-specific; hardware upgrades benefit all workloads uniformly, hence large enterprises set narratives favoring hardware.
  • Traditional algorithm R&D is riskier for investors due to uncertain timelines and outcomes.
• Transition to “algorithm mining”: John’s thesis
  • We are entering an era of automated algorithm discovery where throughput and predictability improve—analogous to industrialized gold extraction versus historical nugget-hunting.
  • Efficient verification paired with scalable search can yield more consistent ROI and unlock investability for algorithmic innovation.

TIG’s core: incentive-aligned, open algorithm markets

• Sparta’s observation: TIG feels like an “obvious missing piece”—a globally distributed mechanism to accelerate algorithm development.
• John’s clarification:
  • Building TIG to be resistant to gaming was non-trivial; that difficulty explains why similar systems may not have emerged sooner.
  • TIG does not itself generate algorithms; it provides the incentive framework for people, machines, or agents to submit algorithms that beat current baselines and then makes them open, verifiable, and monetized via rewards.
• Strategic concern: Centralization risk
  • Sparta: Fears a government/corporate monopoly on algorithm development.
  • John: AI-driven algorithm search removes human bottlenecks; better algorithms beget faster discovery of even better ones—this can become a quasi-military advantage (strategy, intelligence), attracting nation-state interest.
  • Open, decentralized dissemination is essential to prevent lock-in and “dark futures.” TIG exists to keep algorithmic advances open and broadly accessible.

TIG and Bittensor (TAO): relation, non-competitiveness, and collaboration paths

• Sparta’s thesis: TIG and TAO are synergistic, not competitors. Speculates TAO Foundation or certain subnets could even become TIG Treasury companies in the future (long-term possibility, not imminent).
• John’s detailed view:
  • TIG as a Bittensor subnet? Popular question. John plans an article explaining why TIG is not a subnet:
    • Specialization: TIG focuses on problems solvable by algorithmic methods that are hard to compute but efficient to verify. Architecture is optimized for this class; TIG is far less general than Bittensor.
    • Anti-sybil/security model:
      • “Proof of Work” in crypto context is an anti-sybil mechanism (e.g., Bitcoin’s hash power distribution makes capture prohibitively expensive).
      • “Proof of Stake” (PoS) secures via staked token majority constraints.
      • Bittensor uses “proof of work” in the literal sense for tasks, but its anti-sybil security going forward is PoS (subject to OpenTensor’s technical review). TIG, by contrast, can be secured via Proof of Work in the traditional, network-security sense because TIG’s proofs are efficiently verifiable.
      • Conclusion: Different objectives, different security models; TIG’s PoW is preferable for its protocol.
  • Collaboration directions (to be revealed later):
    • Algorithm mining synergy: TIG can incentivize algorithm breakthroughs; Bittensor is well-suited to cultivate and train specialized AI models/agents that search for better algorithms (e.g., “Alpha Evolve” style agents). The mining activity would operate external to TIG’s core protocol but complement TIG’s incentives and market.
• Sparta’s interpretation: Tools like “Alpha Evolve” won’t kill TIG; they can feed algorithms into TIG, accelerating improvement rates. TIG’s incentive layer and open dissemination compound the benefits of agent-driven algorithm search.

Governance, ethics, and the case for decentralization

• Sparta’s concern: Accumulation of power by surveillance-capable firms (Palantir, Oracle), and late-stage capitalism dynamics centralizing both capital and IP.
• John’s perspective:
  • Better algorithms could enable surveillance regimes to prohibit independent AI research and enforce it. Without open dissemination, a monopoly could be locked in for a very long time.
  • Decentralization is necessary at scale: Prevents embargo/protectionism on mathematics and guarantees distribution, akin to Bitcoin ensuring unstoppable value transfer.

Academic engagement and credibility

• Sparta: Notable that leading academics (e.g., Thibault for vehicle routing) choose to work through TIG; signals legitimacy and attractiveness to top researchers. More are engaged (some undisclosed).
• John:
  • Academics can monetize unique algorithmic contributions openly via TIG without trade-offs (e.g., joining closed corporate labs, losing ability to publish/collaborate).
  • Overcoming crypto skepticism: Once the mechanics and necessity of decentralization and tokens are understood, academics recognize TIG couldn’t work any other way.

Optimizer Challenge: scope, importance, and launch timing

• Sparta’s setup: Positioned as one of TIG’s most important challenges; launch targeted for next week (following extensive testnet trials over recent months).
• John’s details:
  • Focus: Neural network training optimizers (minimize loss functions). This is a fundamental algorithm in modern AI.
  • Adam optimizer: One of the most highly cited scientific papers in history, underscoring optimizer importance.
  • Lead researcher: Dr. David Liu (ex-Meta; Cambridge PhD). Finding a new/better optimizer was arduous and took longer than expected—reflecting the high difficulty.
  • Clarification: Not “model optimization” in the vague fine-tuning/tweaking sense. This is the training optimizer used before model weights exist—a core tool for creating the network.
  • Impact: A superior optimizer is likely among the most monetarily valuable algorithmic problems globally. Launch marks a significant milestone for TIG.
• Relation to Bittensor:
  • Many TAO subnets leverage fine-tuned open-weight models (Llama, DeepSeek). Independence requires training from scratch over time.
  • Optimizer breakthroughs must remain open; otherwise, centralized labs with superior private optimizers could outcompete decentralized efforts even against vastly larger distributed compute.

Q&A and community notes

• Adele’s question: About Bob World (leads TAO Quantum Subnets 63 and 48).
  • John: Not familiar; welcomes an introduction. Notes TIG is currently focused on non-quantum computing.
• Raffle: Winners were Magnate and “0”; Sparta will distribute TIG tokens via DM.

Collaboration with ResearchCoin (Brian Armstrong)

• Publicly disclosed activity:
  • ResearchCoin posted a bounty to fund an external review of TIG’s Boolean SAT challenge (an area TIG wanted outside expert validation).
• Ongoing discussions:
  • A more substantive collaboration is in progress but under NDA; details cannot be shared yet.

Key takeaways

• Conferences: Token 2049’s side events produced meaningful engagements; TIG’s narrative resonates, prompting long, high-signal discussions.
• Algorithms vs hardware: Algorithmic advances are at least as impactful as hardware; automated algorithm mining will improve predictability and investability.
• TIG’s role: TIG is an incentive layer and open market for verified algorithmic improvements; it doesn’t generate algorithms itself but ensures access, monetization, and openness.
• Synergy with Bittensor: Non-competitive and complementary; differing security models and protocol specializations. Joint paths likely involve agent-driven algorithm mining feeding TIG’s open incentives.
• Ethics: Centralization dangers are acute; open, decentralized dissemination of algorithms is essential to avoid surveillance-locked monopolies.
• Academia: TIG is attracting top researchers by enabling open publication and monetization; combats the typical corporate trade-offs.
• Optimizer Challenge: A flagship, high-value foundational AI milestone; launching next week after testnet. Success here strongly benefits decentralized AI ecosystems, including TAO subnets.
• Partnerships: Early tangible collaboration with ResearchCoin (external reviews), with more under NDA.

Next steps and upcoming events

• Optimizer Challenge mainnet launch: Next week, following months of testnet hardening.
• John’s article: Planned deep dive on TIG vs Bittensor security models (PoW vs PoS anti-sybil), specialization, and why TIG isn’t a subnet—OpenTensor Foundation will review for technical accuracy.
• Sparta’s explainer: A forthcoming accessible breakdown of the technical article for broader understanding.
• Weekly session: Next Space tentatively Wednesday.