Newsletter #5
01 · From the blog
What we've been writing about
When prompts are not enough: a field guide to AI reliabilitya 35-turn AI game with a key choice on turn three. Pure prompting failed two-thirds of the time on one branch. Multi-agent with code-enforced state kept the choice through to the ending.
Writing code was never the bottleneckAI compresses coding. When review, testing, and release stay slow, faster coding just creates longer queues downstream.
Claude Certified Architect (Foundations) exam: study guide and how I passedthe four Anthropic Skilljar courses, what mock-exam score predicts readiness, and the proctoring traps worth knowing about before you sit down.02 · Open source
ai-agent-subsystem
- ai-agent-subsystem a new public release. Run autonomous coding agents as first-class Kubernetes resources so they slot into the infrastructure you already operate.
03 · Interesting reads
Five things worth your time
- GitLab launched Orbit, a code knowledge graph for AI agents same direction we've been taking with Lore.
- deepreinforce-ai released the Ornith-1.0 collection on Hugging Face open-source coding LLMs from 9B to 397B parameters, 1.6M+ downloads already.
- Ben Rycroft on why even JPMorgan is priced out of frontier inference DeepSeek V4 Pro hits 79% of Claude Opus 4.8's benchmark score at 5% of the cost.
- A hand-cranked AI inference box The Register's tongue-in-cheek angle on AI energy use.
- Alibaba open-sourced their AI code review CLI hybrid deterministic + LLM architecture, reportedly ~1/9th the token cost of general-purpose agents at higher precision.
04 · Events
Three events lined up
Our AI Native Netherlands community has grown to more than 1,600 members in just over a year and a half. We're grateful for everyone that made it possible, and we've got some cool events lined up.
Pattern Cards Workshop
Amsterdam, September 8. An in-person, hands-on session using our 119-card AI pattern cards deck (a shared vocabulary for mapping AI Native architecture and strategy decisions). Details and register
July 30 · AmsterdamStudytube x re:cinqCo-hosted meetup with Studytube. Our CEO Pini Reznik is speaking on our AI Adoption Ladder framework, on how enterprises move from early AI experimentation to organisation-wide adoption.RSVP →
October 27 · AmsterdamAWS x re:cinqCo-hosted meetup with AWS.RSVP →
05 · Training
Public agentic coding training
September 10–11, Amsterdam. Only 4 early bird seats left before regular pricing kicks in, and the cohort is close to full.
Reserve one of the last seatsReply if there's something here you want to dig into, push back on, or ask about. We read every message that comes back.
The re:cinq team
{think ai native}