From prompt to synthesized intelligence — in one pass
Creative Konsoles
creativekonsoles.com · github.com/papjamzzz
Quick Start — 5 Steps
1
Add Your API Keys
Click BYOK in the left panel. Hit Set on each model row and paste your key. Each key stays in your browser session only — never stored, never sent anywhere except the model's own API.
Bring Your Own Keys
2
Set the Faders
Drag each fader to dial in your synthesis profile. High Cost = more context passed between models. High Latency = allow deeper processing. Agreeability pushes models to build on each other vs. diverge. Reliability weights stable models higher.
3
Type Your Prompt
Enter any question, brief, creative challenge, or problem in the Prompt Input box. You can also drag a .txt file onto the input or use ⌘O to open a file. Toggle a System Prompt via ⌘/ to pre-set model behavior.
4
Synthesize
Press ⌘↵ or click the green SYNTHESIZE button. Watch the chain run live — GPT-4 fires first, output flows to Claude, then Gemini, Mistral, and Grok. Each model reads the previous. Press ESC to cancel mid-run.
5
Read the Output
The Synthesis tab shows the final output. Switch to Agreement to see where models converged. Switch to Delta to see transformations at each step. Use ⌘C to copy. ⌘K clears all.
Controls Reference
Fader Bank — Synthesis Profile
Cost
Controls how much context is passed between models. Higher = richer chain, more tokens consumed, slightly higher API spend.
Latency
Allows each model more generation time. Higher = deeper, more considered outputs. Lower = faster responses, more concise.
Agreeability
How much each model builds on the prior vs. challenges it. High = convergent synthesis. Low = adversarial chain, maximum divergence.
Reliability
Weights more stable/deterministic models in the scoring. High = prefer consistency. Low = allow more creative deviation.
Analysis Knobs — Click to Activate
Agree Detect
Flags passages where 3+ models used near-identical phrasing.
Consensus Map
Visual cluster of where all 5 models agree vs. diverge.
Delta Analysis
Tracks what each model added, removed, or reframed from the prior step.
Confidence Score
Aggregate certainty across all 5. High = strong consensus. Low = high divergence.
The Synthesis Chain — 5 Models, 1 Pass
GPT-4
gpt-4o-mini · OpenAI
Position 1 — Seeds the chain
Claude
claude-haiku-3-5 · Anthropic
Position 2 — Refines + reasons
Gemini
gemini-2.0-flash-lite · Google
Position 3 — Expands + grounds
Mistral
mistral-small-latest · Mistral AI
Position 4 — Challenges + sharpens
Grok
grok-3-mini · xAI
Position 5 — Final synthesis
BYOK — Bring Your Own Keys
5i never stores your API keys. They live in your browser session only and are sent directly to each model's official API endpoint — nowhere else. Click BYOK in the left panel, then hit Set on each model to enter your key. Keys are cleared when you close the tab.
Randomize to explore. Hit ⌘R before running to discover synthesis profiles you wouldn't dial in manually. Every combination produces a different chain behavior.
◈
Use the System Prompt. Pre-load a persona, constraint, or domain brief. Every model in the chain inherits it — forces consistent voice across 5 different intelligences.
◈
Drop a .txt file in. Drag any text document onto the prompt box. Brief documents, research notes, code snippets — all valid inputs for chain synthesis.
◈
Watch the Monitoring panel. THREAD shows live CPU load. LAST RUN shows exact cost per synthesis. SESSION total tracks your spend. Typical run: under $0.003.
◈
Toggle models in the chain. Click any model in the Synthesis Chain strip to disable it. Run with 3 or 4 models for speed and cost experiments.