Selene Blog
Borrowing modern reading UX patterns from 2025-2026 blogs: clear cards, useful metadata, side navigation, and media-rich storytelling with practical implementation detail.
Apr 21, 2026 • 7 min read
Most AI coding bills look unpredictable because the pricing is built out of synthetic credits on top of real tokens. Here is a four-variable cost model, verified Anthropic and OpenAI rate cards, and an honest monthly bill for a five-developer team — $368 uncached or roughly $150 with prompt caching, not $3,000.
Apr 21, 2026 • 4 min read
Claude, Devin, Jules, Goose, Harvey — every AI agent people actually call by name went out the door with a human handle, not a version string. A short, warm read on why names matter, a lineup of real named agents in production, and three rules we now use when we ship a new one.
Apr 19, 2026 • 3 min read
Every AI coding vendor eventually re-prices its plans — Augment Code is a recent example, and the same gravity pulls on everyone else. Here’s why subscription pricing is structurally unstable, what an MIT license actually guarantees, and how Selene’s BYOK + self-host architecture shrinks the surface where promises can break.
Apr 19, 2026 • 7 min read
A team swapped Augment Code at ~$3,000/month for Claude Code at $200/seat and saved 90%. The three AI coding pricing models — credit-metered, flat-rate and BYOK self-host — a five-developer month priced at provider list rates, and a migration checklist that works coming off any bundled tool.
Apr 17, 2026 • 9 min read
A transcript-with-commentary of a real code review run on Selene. Five subagents, five lenses, one small delegation-lifecycle commit, dispatched as a single parallel batch. We walk through the fan-out, the findings clustered by theme, and the synthesis that turned them into concrete follow-ups.
Apr 17, 2026 • 7 min read
A short, honest note on how AI coding tool pricing evolved over the last eighteen months — and the five architectural decisions (open source, BYOK, model-agnostic per-task models, open MCP tool protocol, real multi-agent delegation) that we made to give Selene users a platform they can depend on for years.
Mar 28, 2026 • 6 min read
Walk through setting up Selene from scratch, configuring vector search, and building a complete Swift card game from a single prompt.
Mar 20, 2026 • 4 min read
Selene now plays lightsaber sounds during agent work — ignitions on session start, swings on tool use, clashes on failures. Plus voice overlay and multi-agent demos.
Mar 14, 2026 • 6 min read
This was my first real SWE-bench Lite pass with Selene, not a polished rerun. I used Claude Opus 4.6 in non-thinking mode, kept the default Selene agent, ran tasks sequentially, and still landed at 182 resolved out of 300.