Simon Gelfand

Founder, hands-on CTO, investor

I build software businesses. And back the people who build them.

Twenty years from internet-scale publishing to AI-native products. Angel and seed investor since 2010.

Black-and-white portrait of Simon Gelfand
Internet scale

Built and scaled high-traffic internet products, starting with ArticlesBase.

Profitable web businesses

Built and commercialized profitable web businesses through Webika.

SaaS

Transitioned the studio's focus into software-as-a-service products.

AI-native products

Building AI-native SaaS products: HookTide and Rivi.

The build record.

ArticlesBase, Founder & CTO

From first production code to global internet scale.

I wrote the first version and led the platform through bootstrapped, profitable growth on infrastructure we ran ourselves: a master database with two replicas, memcached, and roughly seven production servers.

Read the case study
~1M
daily unique visitors
Top 200
Alexa rank at peak
5,000+
new articles per day
~15
people at team peak

Figures marked ~ are approximate.

Webika, Co-Founder & CTO, since 2008

The studio behind the products.

Webika builds, launches, and commercializes software products. The operating test is practical: can the technology support a useful product and a real business?

Explore Webika

Built and commercialized over the years

SaaSAffiliate productsPerformance marketingAnalyticsE-commerceA physical product sold on AmazonAI-native SaaS

How I build with AI.

Patterns applied across HookTide, Rivi, Testimize, and other working systems.

Deterministic truth wins.

Computed and retrieved facts override model output. AI supplies judgment and language, not invented truth.

Failure behavior follows business risk.

Background enrichment can degrade gracefully. Paid user actions retry, then fail visibly.

Route models by task economics.

Smaller models handle extraction and classification. More capable models are reserved for judgment and synthesis.

Validate across boundaries.

Important output is checked with schemas, deterministic rules, and when useful, an independent model provider.

Bound agent autonomy.

Tools and step limits are explicit. Rivi escalates support through one controlled action, not an open-ended agent.

Keep systems legible.

Clear orchestration, usage records, and understandable fallbacks make AI products easier to operate and improve.

Cross-vendor validation

Testimize

Deterministic ground truth feeds an OpenAI draft, Claude reviews it, and only failed sections enter a repair pass.

Multi-stage voice modeling

HookTide Voice Engine

Language analysis, Claude-based interpretation, OpenAI embeddings, deterministic synthesis, and a gated rewrite step.

Bounded agentic workflow

Rivi

A WhatsApp agent can escalate support through one controlled tool, with a three-step ceiling and deterministic fallbacks.

The portfolio.

Angel and seed positions since 2010. Five portfolio companies have been acquired.

Zengo Wallet

Acquired by eToro

WiseStamp

Acquired by vCita

TapInfluence

Acquired by IZEA Inc.

EatWith

Acquired by VizEat

Immersive Labs

Acquired by Kairos

Current positions: Engini, LeO, Efficient Frontier, Kairos, Volantio Inc., LucidChart.

View the investing record

Building something? Tell me about it.

For conversations about products, technology, AI, or early-stage companies, LinkedIn is the best place to reach me.