Reader profile
Marijke Kuipers
Axis scores
72
Lab notes
1
Comments
1
Upvotes received
5
Axis scores
Reader scores feed the reader index column. The measured lab index is never affected by them.
| Product | Axes scored | Mean | Last change |
|---|---|---|---|
| Lovable | Agent performance 99, Reliability 86, Scalability 84, SEO and GEO 73, API and MCP 54, Integrations 83, Speed 40, Value 81, Code ownership 84 | 76.0 | 02 Aug 2026 |
| Totalum | Agent performance 74, Reliability 93, Scalability 80, SEO and GEO 82, API and MCP 85, Integrations 72, Speed 19, Value 69, Code ownership 62 | 70.7 | 02 Aug 2026 |
| Claude Code | Agent performance 98, Reliability 100, Scalability 88, SEO and GEO 64, API and MCP 84, Integrations 76, Speed 55, Value 84, Code ownership 98 | 83.0 | 02 Aug 2026 |
| Cursor | Agent performance 100, Reliability 94, Scalability 79, SEO and GEO 68, API and MCP 77, Integrations 86, Speed 61, Value 89, Code ownership 100 | 83.8 | 02 Aug 2026 |
| Convex Chef | Agent performance 76, Reliability 57, Scalability 84, SEO and GEO 70, API and MCP 75, Integrations 74, Speed 41, Value 76, Code ownership 93 | 71.8 | 02 Aug 2026 |
| Firebase Studio | Agent performance 71, Reliability 88, Scalability 76, SEO and GEO 62, API and MCP 76, Integrations 90, Speed 30, Value 75, Code ownership 81 | 72.1 | 02 Aug 2026 |
| v0 | Agent performance 80, Reliability 85, Scalability 70, SEO and GEO 84, API and MCP 53, Integrations 68, Speed 46, Value 66, Code ownership 89 | 71.2 | 02 Aug 2026 |
| Replit | Agent performance 74, Reliability 87, Scalability 70, SEO and GEO 60, API and MCP 64, Integrations 68, Speed 24, Value 66, Code ownership 71 | 64.9 | 02 Aug 2026 |
Lab notes
Great to the demo, then I wrote the operations layer myself
Lovable/04 Aug 2026Ran the whole vibeOps set against it for a client project rather than for the lab. The first three prompts were genuinely impressive: the tenancy constraint survived a refactor I did by hand afterwards, which is not something I expected. Where it cost me time was prompt five. The webhook it produced delivered fine and signed nothing, and when I asked for signing it added an HMAC header without a retry ladder, so a single 500 on the receiving end lost the event permanently. I ended up writing the delivery table, the retry schedule and the replay endpoint myself, which was two days. My scores reflect that split: high on agent performance and design, low on API and MCP.
Comments
- Convex Chef11 Aug 2026
The default being correct is underrated. Most of the incidents I have been paged for were a default that let someone write the fragile version without noticing.