goodwaves.news · 2026 · Seed Round Open

The world's first
positive-only
social network

No doom. No outrage. Only waves of good news shared by real people — governed by agentic AI, built on radical joy.

Scroll
€75MARR target year 3
97%Agentic AI moderation
<200msHigh-confidence approval
€0CAC at launch
GoodWaves logo rising from the ocean at night
The logo rises
Three phones emerging with GoodWaves UI
The product emerges
Globe with GoodWaves platform
The world connects
GoodWaves on devices

Beautiful.
Positive.
Addictively good.

Five core features engineered to create the same habitual engagement as Duolingo and Strava — applied to good news instead of language learning.

Wave Feed — location-based positive news
Wave Energy & 7 Tide Levels gamification
Wave Map — live geo-pinned good news
Share Cards — every share earns Wave Energy
GoodWaves Plus — €4.99/mo subscription
GoodWaves AI globe

Agentic AI Moderation Pipeline

A stateless classifier is a bouncer with a rulebook. Our agentic system is an experienced editorial board with memory, multi-step reasoning, and tools.

Latency design principle

The pipeline is deliberately tiered for speed. Posts scoring ≥72 with high confidence auto-approve in sub-200ms — imperceptible to the user.

Steps 3–5 only activate for the ~15% in the ambiguous 55–72 band. 97% of posts resolve before the user finishes reading the success confirmation.

Live pipeline · EU-hosted · 27 languages
High-confidence post
<200ms
Ambiguous 55–72 band
~1.5s
Edge-case review
<3s
Human governance
3% of posts
Traffic split
■ 82% auto-approve ■ 15% async ■ 3% human
Step 01
⚙️
Ingestion & Pre-processing
Always active · every post

Strips metadata, normalises text, detects language across 27 European languages. Posts with images route in parallel to a vision model. Logs full pseudonymised context into short-term memory — purged after 90 days per GDPR.

NLPVision model27 languagesGDPR Art.5
Latency<10ms
Step 02
🧠
Primary Sentiment Agent
Instant approval path

Fine-tuned transformer (Mistral 7B / Llama 3 class) evaluates four dimensions: factual positivity, emotional tone, social impact framing, and authenticity signals. Produces Joy Score 0–100 with confidence interval. ≥72 + high confidence: auto-approves instantly.

Joy Score 0–100Mistral 7BConfidence interval
Fast path<200ms
Step 03
🛡️
Adversarial Check Agent
Anti-gaming · sarcasm detection

Separate model trained on 800K irony-annotated adversarial examples. Hunts for positive framing on genuinely negative events, sarcasm with positive surface language, and toxic positivity — dismissiveness of real suffering. Can modify the Joy Score or escalate.

800K examplesSarcasm modelIrony detection
Activates for~15% of posts
Step 04
🌐
Context Agent
Ambiguous content only

Activates only for the ~15% in the 55–72 band. Pulls pseudonymised post history, current regional trending topics, and related posts. A verified journalist posting a nuanced recovery story receives different treatment than an anonymous account with unusual patterns.

Post historyRegional trendsAccount signals
Async processing~1.5s avg
Step 05
Decision & Action Agent
Hallucination guard · constrained gen

Decline explanations use constrained generation — selecting from pre-validated, brand-reviewed templates. Free-form generation is never used for user-facing messages. Eliminates hallucinated reasons and satisfies DSA Article 17(3) transparency requirements for automated decisions.

Constrained genDSA Art.17No free-form AI
Decline explainability100% templated
Step 06
👁️
Human Governance Layer
3% of posts · EU AI Act Art. 14

3% of posts reach a Trust & Safety analyst — an analyst whose decision feeds immediately as a labelled training example. Every review makes the AI smarter. The EU AI Act requires this layer for consequential automated decisions. Regulatory compliance and model training simultaneously.

EU AI ActTraining signalPolicy governance
Coverage3% intentional
Step 07
🔄
Appeals Loop
<0.1% · continuous improvement

Users contest declines via a structured appeals interface. Senior Trust & Safety rulings become high-weight training signals. Consistent appeal patterns trigger model audits. User objection becomes product improvement — turning human oversight into a training asset.

High-weight signalModel auditContinuous learning
Volume<0.1%
Score <55
Fast Decline
Constrained template · policy clause cited · no free-form AI · compassionate reframe offered
Score 55–72
Async Review
Steps 3–5 activate in parallel · context agent · adversarial check · max 3s wait
Score ≥72
Auto-Approve
High-confidence · sub-200ms · Wave Energy awarded · story enters feed instantly
DECLINE
ASYNC
APPROVE
05572100

Moderation is not a cost centre.
It is the revenue intelligence layer.

Posts scoring ≥85 ('High Energy') are auto-flagged as Local Spotlight candidates. The same AI that protects the feed surfaces premium ad inventory. At 30M MAU, own-model GPU hosting reduces inference cost a further 80%.

ScaleMAUAgentic AI / yrHuman-only equiv.vs HumanSaving
Launch Y1250K~€8K~€320K
97%
Grow Y23M~€48K~€1.8M
97%
Scale Y315M~€180K~€7.5M+
97%
Global Y4+30M~€280K*~€15M+
98%*

* At 30M MAU, own-model GPU hosting reduces inference cost 80% vs API pricing.

Better moderation → better content → more revenue → better model

🛡️
Better moderation
Higher Joy Score precision
Better content
Higher quality feed
📍
Spotlight value
Premium inventory surfaced
💶
More revenue
Plus + Spotlights + API
🧠
Better model
Training reinvestment
Stream 01
GoodWaves Plus
€4.99/mo · €39.99/yr · €8.99 Family

2× Wave Energy, Story Boost into Global Feed, unlimited radius, analytics, exclusive themes. 3% conversion at launch → 5% at scale. AI identifies High Energy stories for Plus Boosts. Strava benchmark: ~4.5%.

📍
Stream 02
Local Spotlights
€49/wk · €149/mo · €790/mo

Businesses sponsor city-level positive story placements — all pass the positivity filter. B2G layer: municipalities pay €490/mo for a Verified Civic Channel. 50 cities Year 2 = €294K/yr recurring B2G ARR.

🔌
Stream 03
White-Label API
Enterprise pricing

Schools, hospitals, corporate wellness teams and media outlets embed curated GoodWaves feeds in their own apps and intranets. B2B2C distribution at zero additional content cost.

🎉
Stream 04
Events & Merch
Variable

GoodWaves Joy Summits, local meetups, branded merchandise. Community revenue that deepens loyalty and generates massive organic reach with zero advertising spend.

€75M ARR by 2028

2026 · Launch + Grow
€4M
Year 1 ARR
MAU3M
Plus (3%)90K · €3.6M
Spotlights€400K
API
2028 · Global
€75M
2.6× growth
MAU30M
Plus (5%)1.5M · €66M
Spotlights€5M
API€4M

There is no GoodWaves.
That's the opportunity.

FeatureGoodWavesTwitter/XFacebookRedditPositive News sites
Positive-only content
Social / community
Gamification / rewards
Location-based feedpartialpartial
Agentic AI moderation
EU data sovereignty
Ad-free subscription
Age safeguarding (16+)

From launch to
global good waves

Q2
Q2 2026 · Launch
Europe
  • iOS + Android apps
  • AI filter live
  • 3 launch cities
  • Founding Creators programme
  • 250K early adopters
Q4
Q3–Q4 2026 · Grow
North America
  • North America launch
  • GoodWaves Plus live
  • Business Spotlights
  • 50 municipality partnerships
  • 3M MAU target
27
2027 · Scale
Global
  • API + white-label
  • 20 markets
  • Joy Summit events
  • Own-model GPU hosting
  • 15M MAU target

Built by people who believe
good news changes the world

IN
Chief AI Architect & Generative Media Futurist
Ivaylo Nikolov

AI strategist and generative media specialist focused on the intersection of artificial intelligence, content, and human wellbeing. Ivaylo brings a practitioner's perspective to GoodWaves — translating cutting-edge AI capabilities into real-world social media impact across European markets.

VC moat: GENERATIVE AI DIFFERENTIATION · Ivaylo's practitioner-level command of generative AI and content systems means GoodWaves doesn't merely use AI as a tool — the product is built around AI as a first-principles competitive advantage. His rare specialisation at the intersection of AI and human wellbeing directly powers the Joy Score Engine, the positivity filter architecture, and the platform's core thesis: that algorithmic design can actively improve mental health at scale. That is not a feature. That is a structural moat competitors cannot replicate without rebuilding from zero.
OG
CEO & CTO
Orvar Geir Fridriksson

Multi-level computer vision scientist, AI Agentic Researcher and Analyst. Serial entrepreneur across 10+ companies in breakthrough technology across the USA, Iceland, Germany, Italy and Spain.

VC moat: SCALABLE INFRASTRUCTURE · The agentic moderation pipeline is the CTO's prior art — not a roadmap item.

Seed round · 2026

€6.46M

We are raising a €6.46M seed round to build the technology, acquire the first 250,000 Europeans, and prove the Zero CAC flywheel.

50% Product & Engineering
25% Growth & Marketing
15% Compliance & Legal
10% Partnerships & BD

goodwaves@goodwaves.news · Ivaylo Nikolov · Orvar Geir Fridriksson