Sinther

Inspiration

Not everyone gets lucky with teammates. In most hackathons — and in life — the quality of your decisions depends heavily on who's in the room. Some people think strategically, others spot risk, some dream creatively, others read the human side. Rarely does one person — or even one team — cover all six.

We were inspired by Edward de Bono's Six Thinking Hats and Belbin's Team Role theory, both of which prove that optimal decisions require six distinct cognitive perspectives: logic, risk, optimism, creativity, empathy, and synthesis. The best teams in the world aren't the smartest — they're the most cognitively diverse.

So we asked: what if everyone had access to that dream team, on demand?

That question became SINTHER.

What it does

SINTHER is a chat-based AI decision-making platform where every idea, proposal, or dilemma is analyzed by 6 specialized AI agents — each embodying a scientifically backed cognitive role:

• Chanky (The Strategy Architect) — objective logic and structured planning
• Vizi (The Risk Mitigator) — pessimistic realism and worst-case analysis
• Kavi (The Value Optimiser) — constructive optimism and hidden upside
• Sri (The Innovation Catalyst) — lateral thinking and creative breakthroughs
• Jupi (The Culture Guardian) — human empathy and cultural awareness
• Mars (The Conductor) — metacognitive synthesis of all perspectives

After all agents respond sequentially, SINTHER auto-generates a Summary Message that distills every viewpoint into one clear, actionable verdict. Users can toggle agents on/off, adjust a Rationality Slider to control response strictness, and save full chat history when logged in.

How we built it

SINTHER was built during the Medo Hackathon as a full-stack web application. The frontend is a clean, dark-themed chat interface with real-time sequential agent responses. Each agent is powered by a distinct system prompt engineered to faithfully embody their cognitive archetype. The settings page lets users configure which agents participate and how strictly they respond. We also built guest mode (no signup, session-only) and a registered user mode with persistent chat history.

Challenges we ran into

The hardest challenge was prompt engineering at scale — making six AI agents feel genuinely different from one another, not just surface-level variations. Vizi had to feel truly pessimistic without being unhelpful. Sri needed to be creatively lateral without being incoherent. Getting Mars to synthesize five prior responses into something non-redundant required careful context design.

We also had to carefully manage sequential response flow — ensuring agents responded in order, each response visible before the next begins, without race conditions or UI jank.

Accomplishments that we're proud of

We're proud that SINTHER doesn't just feel like six AI agents — it actually reasons differently across all six. The Summary Message in particular felt like a genuine breakthrough: watching six opposing perspectives collapse into one clear recommendation is exactly the kind of cognitive clarity we set out to build.

We're also proud of building something grounded in real science — not just another AI chatbot, but a product with a theoretical backbone that justifies why six agents, why these roles, and why this order.

What we learned

We learned that the hardest part of multi-agent AI isn't the technology — it's the character design. Each agent needed a distinct voice, reasoning style, and epistemic stance. We also learned how much prompt structure matters: small changes in system prompts produced wildly different decision quality.

And we learned something personal: the very problem SINTHER solves — not having diverse thinkers around you — was something we felt building it. SINTHER is the teammate we wished we had.

What's next for Sinther

• Memory across sessions — agents that remember your past decisions and evolve their advice
• Team mode — multiple users collaborating with shared agent sessions
• Domain specialization — agents tuned for specific fields (legal, medical, financial, product)
• Decision tracking dashboard — visualize how your decisions have played out over time
• Voice input and audio responses — for on-the-go decision support
• API access — so developers can embed SINTHER's cognitive engine into their own products

Built With

• medo