Tech • AI • Robotics • Game
Base 44 vs Lovable, Convex backend, Tapit Store launch
Briefing
Base 44 pressures Lovable
Base 44 emerged as the stronger option for structural changes in larger, lived-in apps after extended production-style use. The comparison centered on a harder test than first-run prototyping: schema changes, existing users, populated databases and the risk of breaking live workflows. Lovable still stood out for fast starts, polished first versions and lightweight internal tools. The split increasingly looks like one of lifecycle fit rather than a single winner.
Lovable keeps its speed edge
Lovable retained a clear advantage for landing pages, prototypes and smaller tools that can often be generated in one or two prompts. A cited survey of more than 200 developers put satisfaction at roughly 85% for landing pages, 80% for visual prototypes and 65% for simple internal tools. Its large community also appears to reduce friction when debugging or finding workarounds. That makes Lovable hard to dislodge for teams optimizing for launch speed over long-term refactoring.
Lovable growth resets expectations
Lovable's commercial scale remains a major part of the story. In December 2025, the company raised $330 million at a $6.6 billion valuation in a round led by Capital G, with backing from Menlo Ventures, N Ventures, Salesforce Ventures and Databricks Ventures. It reportedly hit $100 million ARR within 8 months and passed $200 million four months later. That growth suggests investor conviction that AI app builders can become foundational software platforms, not just prototyping tools.
Convex pitches backend in one box
Convex is being positioned as a bundled backend layer that replaces the usual patchwork of database, API, auth, storage and real-time sync tools. The appeal is operational simplicity: one platform to store data, run server logic and keep clients synchronized. For AI-assisted development, that packaging matters because fewer moving parts means fewer integration failures for generated code. The product is effectively betting that backend consolidation is the next productivity unlock after frontend code generation.
Convex leans on reactive sync
A key differentiator for Convex is its reactive model, where changes made in one client propagate automatically to others in real time. The platform organizes server behavior around queries, mutations and actions, giving apps a tighter default structure than ad hoc backend stacks. That design aims to reduce stale state, manual refresh logic and cross-device inconsistencies. In practice, it targets a persistent pain point for collaborative and multi-device products that AI coding tools often struggle to wire correctly.
Tokens become the real meter
Tokens remain the unit that models actually process and bill against, not human-readable words or sentences. Because tokenization varies by model, the same prompt can be split differently by systems such as GPT-4o, with direct consequences for cost and latency. A rough rule of thumb put 431 characters at about 104 tokens, though language and vocabulary shift the ratio. For teams building agentic products, token awareness is increasingly becoming a product-design constraint rather than a niche implementation detail.
Context windows hide reliability tradeoffs
The briefing on context underscored that long conversations are not just more expensive; they also become less reliable as working memory fills up. Input and output tokens are both billed, and one cited example for Sonnet priced usage around $2 for input and $10 for output. That asymmetry means verbose prompts and sprawling responses can degrade margins quickly. The broader implication is that prompt discipline, retrieval design and memory management are now core engineering issues for AI coding products.
Tapit Store chases creator revenue
Tapit is expanding from links, QR codes and conversion tools into direct commerce with Store, a simplified selling product for creators. The target user is visible online but under-monetized, with the tool focused on selling PDFs, courses, tickets, calls and other digital goods through an easy checkout flow. In parallel, early ad-creation tests suggested Sidence 2.5 and Kling 4.0 generated the strongest first-pass marketing clips. Together, the move points to a tighter coupling between creator monetization software and AI-generated promotional assets.
Videos covered
700 Days on Lovable VS 31 Days on Base44
- •Why the comparison matters
- •Where Lovable performs well
- •Lovable’s scale and pricing strengths
Pourquoi Claude te coûte plus cher à chaque nouveau message
- •AI reads tokens, not sentences
- •Tokenization varies by model and language
- •Why tokens matter for quality and cost
Convex : l'alternative à Supabase moins chère et plus performante
- •What Convex replaces
- •How the app model works
- •Reactive synchronization
Comment avoir un SECOND CERVEAU parfait sur Claude ? (le système simple)
- •A reaction against overbuilt systems
- •The core idea is a file-based RAG
- •Why plain text beats a database here
On a testé TOUS les modèles d’IA vidéo pour voir lequel est le MEILLEUR
- •A monetization push beyond links
- •Store targets creators who are visible but under-earning
- •A simple checkout for digital products