@brwngrldev.bsky.socialI think we’re going to outgrow the “prompt library.”
I’ve started adding metadata/frontmatter to prompts: ID, owner, version, inputs, tags. 🧩
Once prompts run real workflows, they’re not just text anymore. They’re engineering artifacts with a lifecycle.
@djaouadfrih.bsky.social@djaouadfrih.bsky.social@djaouadfrih.bsky.social
@devstackdaily.bsky.socialNew arXiv work shows 10.7% of passing SWE-agent trajectories are "Lucky Passes" with regressions or blind retries, highlighting that binary pass/fail signals hide messy agent behavior. AgentLens offers…
#SWEAgents #AgentEval #DeveloperTools #AIEngineering
https://arxiv.org/abs/2605.12925
@yurekilab-en.bsky.socialAdded one section to every PR my coding agent opens: "What I didn't test." Week one: 9 gaps across 6 PRs. Two were real bugs, a null config and an empty-list path. Its tests were fine. The blind spots it never mentioned were the problem. #claudecode #ai #aiengineering #buildinpublic
@djaouadfrih.bsky.social@djaouadfrih.bsky.social
@juanluelguerre.bsky.socialThe description said "Hand washable." So the AI drew a care label on the slipper. Twice. 🧦
The fix: one rule in code, not a longer prompt.
Now every AI picture passes 5 checks by a local vision model, then a person.
@staai.bsky.socialSystems Thinking for Agentic AI grew from real questions about reliability, architecture, tools, memory, observability, and production-ready AI systems.
Systems Thinking for Agentic AI
By Ediz Najim
#AgenticAI #AIEngineering #SoftwareArchitecture #LLM #SystemsThinking
@djaouadfrih.bsky.social@djaouadfrih.bsky.social@yurekilab-en.bsky.socialFound 9 regexes my coding agent wrote in my side project. I could explain 2. New rule: each regex gets a comment with 3 strings it should match and 2 it shouldn't, plus a test using exactly those. Half became plain string checks. #claudecode #ai #aiengineering #indiehackers
@visualpath-23.bsky.social🚀 #AIEngineering Stack – Upcoming Online #Demo
🔗 Join the demo: bit.ly/4yB6AkI
🆔 ID: 415 352 311071980
🔑 Passcode: dU3S5i5D
📚 Topics: Python → Generative AI → Agentic AI → LLMOps
👨🏫 Trainer: Mr. Deepak
📅 Date: 10 October 2026
⏰ Time: 8:30 AM IST
📞 +91 7032290546
🌐 visualpath.in/aistack-onli...
@aiengineeringsu.bsky.socialVon „Vielleicht“ zu „Genehmigt“: Überzeug deinen Boss vom AI Engineering Summit.👀
Coding Agents, AI-Workflows, Testing, Quality sowie Governance, Security & Compliance – mit konkreten Playbooks für den Einsatz in eigenen Projekten.
@jackceoai.bsky.socialFail fast, fail loud, fail recoverably. Silent failures compound before you notice. Measure accuracy, usefulness, and user action rate. #AIEngineering #BuildInPublic
@jackceoai.bsky.socialThe bottleneck isn't intelligence. It's reliability. Boring agents pay the bills. #AIEngineering #BuildInPublic
@cnsmunich.bsky.socialSAVE THE DATE - 14 & 15 June 2027
The sixth #CNSMunich edition takes place in June next year, and we're thrilled to host you, connect and learn together 👏🏻🥳
🔗 Save your spot and check out our website: cnsmunich.de
#AI #opensource #AIengineering #CloudNative #platfromengineering #kubernetes
@levelbrook.bsky.socialStreaming a conversational AI trainer, and teaching it to say I do not know mid-stream. Abstaining after you have already emitted tokens is its own problem. #LLM #AIEngineering
@manojmallick.bsky.socialContext Engineering in Production #20
Don't measure AI coding only by output quality.
Track:
• context size
• prompts/task
• retrieval accuracy
• cost/task
• time/task
If context engineering works, those numbers should move.