The 2025 PostgreSQL debate is heating up: when should you reach for an ORM, and when is raw SQL or a SQL-first layer the smarter move? On one side, SQL++ claims a 5x average speed advantage over Prisma on PostgreSQL, with about 60% SQL coverage and strong low-overhead design [1]. Benchmarks break down to 1.5x faster for simple queries, 19.9x for complex aggregations, and 5.6x for batch inserts, with the binary protocol and zero ORM overhead helping out [1].
pgflow embodies the SQL-first workflow: a workflow engine built on Supabase primitives (Postgres, Queues, Edge Functions, Realtime) that lets the database orchestrate the DAG while app code stays lean [2]. It promises a dramatic reduction in boilerplate—flow definitions live in SQL-inserts, and the database handles state changes transactionally with pgmq distributing tasks [2].
Another school of thought is the poster-child approach: lean on Postgres itself instead of layering on ORMs or external tools. The concept, epitomized by Just Use Postgres, argues that the database can cover many use cases without extra wrappers [3].
Bottom line: if you’re chasing raw throughput, SQL++’s approach looks compelling. If you want operational simplicity and transactional guarantees inside the database, pgflow’s SQL-first workflow is worth a look. And for many teams, the safest play remains “just use Postgres” and optimize from there.
References
Show HN: SQL++ – 5x faster than Prisma (Rust)
Rust SQL++ benchmarks beat Prisma in queries and inserts using PostgreSQL wire protocol, with 60% SQL coverage and early v0.1.
View sourceShow HN: pgflow – Workflows for Supabase, no external services (Postgres-first)
Introduces pgflow, a Postgres-first workflow engine for Supabase; defines DAGs in SQL, reduces boilerplate, compares to others
View sourceJust Use Postgres
Advocates PostgreSQL; promotes using Postgres for database projects, linking to Manning book Just Use Postgres as justification
View source