Changelog
What's new in Hyves (Holistic Yield & Validation Engines).
Hyves takes its name
The platform is now Hyves — Holistic Yield & Validation Engines — built by iSYNCSO. Engines that yield and validate, holistically, on one memory.
- Product renamed across the platform, sites, and documentation
- Hex design grammar unified end to end: login, onboarding, and platform share one visual system
- SYNC site guide launched — the agent now shows visitors around and answers questions on the public sites
- New public field manual documenting all ten steps of the method
Self-thinking field missions
Give Fields a goal in plain language — "all clients of a competitor" — and the engine works out the how: which sources to read, which parsers to mint, and which proof counts.
- Goal-driven prospect missions that plan their own research route
- Review-platform mining to surface companies using a named competitor
- People resolved to verified profiles and email, then harvested into your universe
- Results flow straight into the account map — no export step
Durable worker-bee swarm
Missions now decompose into queued child runs with a plan-approval gate, so a slow task can never stall the swarm — and nothing executes before you approve the plan.
- Manager agents decompose a mission and queue specialized workers
- Human approval gate between the plan and any execution
- Scheduler-driven rollup: long jobs finish in the background and report back
- Sequential chains — steps that depend on earlier results wait for them, and failures cascade cleanly
Structured pricing and deeper foraging
The foraging pipeline now reads structured pricing and product detail from the open web, and the research run goes deeper before it scores.
- Structured pricing extraction during the products research stage
- Deeper crawl budget with per-product deep dives
- Proposition red-teaming — the engine argues against its own conclusions before presenting them
- Assumptions and signals carried from the playbook into scoring and enrichment
Deep Intelligence Engine
Two-pass LLM reasoning engine that produces cross-domain business suggestions by analyzing your entire data landscape.
- Hourly intelligence collection across 10 data domains
- Two-pass LLM reasoning (collect → reason) for nuanced insights
- User intelligence profiles computed nightly across 5 dimensions
- Temporal pattern detection for deadlines, seasonality, and trends
- Suggestions cite specific names, EUR amounts, and deadlines
Flow Automation
Visual DAG builder for autonomous outreach pipelines. Build Lumos → Flow → Demo Booked sequences with AI-powered prospect research.
- Visual DAG builder for multi-step automation flows
- Lumos AI research integration for prospect enrichment
- AI-powered personalized email crafting with quality gate
- Reply detection and classification (positive/negative/neutral)
- Autonomous monitoring and self-healing via flow-monitor cron
Voice Mode Optimization
Sub-second voice responses by eliminating server-side TTS and using browser speechSynthesis for instant audio playback.
- Direct LLM call replaces full pipeline proxy chain
- Browser speechSynthesis for zero-latency audio
- Response time reduced from 20+ seconds to sub-second
- Llama-3.3-70B-Instruct-Turbo for faster voice inference
- Client-side conversation history for seamless context
Blueprint Build Complete
Full client onboarding system with email pool management and Shopify integration for e-commerce workflows.
- Client onboarding wizard with guided setup
- Email pool with rotation, warm-up, and health monitoring
- Shopify integration for product sync and order management
- Google Sheet audit trail for all automated actions
Semantic Upgrade Phase 4
Product intelligence, scheduling recommendations, and agent personalization — the final phase of the semantic upgrade.
- Product health scores and margin tracking
- Scheduling recommendations based on behavioral signatures
- Agent personalization via feedback loops and learned preferences
- Usage tracking with adaptive UI based on feature adoption
- Trigger management for third-party event handling
FLUX.2 Image Generation
5-tier AI image generation system with smart auto-selection. Create marketing visuals, product images, and social content from natural language.
- 5 generation tiers from fast drafts to ultra-high quality
- Smart auto-selection based on use case and complexity
- Integrated into Create engine for seamless content workflows
- Batch generation for marketing campaigns