Running AI agents without governance is a risk your team can't afford. Upsun Dispatch™ is the fix. 🤖 Launching at #dotAI2026, don't miss out. upsun.com/dispatch/ #UpsunDispatch #AIEngineering
@upsun.com
@upsun.comRunning AI agents without governance is a risk your team can't afford. Upsun Dispatch™ is the fix. 🤖 Launching at #dotAI2026, don't miss out. upsun.com/dispatch/ #UpsunDispatch #AIEngineering
@mapjournals.bsky.social#KER
Can LLMs design analog circuits like human engineers? A structure-performance reasoning framework maps SPICE netlists to text generation.
Details: https://www.maxapress.com/article/doi/10.48130/ker-0026-0012
#AnalogCircuits #LLM #CircuitDesign #AIEngineering
@amkentech.comHit the invisible wall of connector limits? Most guides leave you there, stuck. We built patterns to extend your automation beyond standard tools. Get the Beyond the Connector guide and keep building. → https://amkentech.com/guide/connector #Automation #AIengineering #MCP
@hackernoon.comA practical guide to spotting over-engineered AI apps, from unnecessary vector databases to multi-agent systems, memory layers, and missing evals. #aiengineering
@yurekilab-en.bsky.socialI kill my coding agent sessions on purpose now. Past a couple of hours, it starts defending old decisions instead of rereading the code. A fresh session with a 10-line handoff note beats 100k tokens of stale context every time. #claudecode #ai #aiengineering #buildinpublic
@yurekilab-en.bsky.socialTrimmed my coding agent's instruction file from ~180 lines to 40. Behavior barely changed. Most rules were patches for mistakes older models made and newer ones just don't. Instructions rot like code — nobody deletes them. #claudecode #ai #aiengineering #buildinpublic
@58bits.comMy current description of SOTA LLMs? They're like savant kids with tunnel vision. Very good once they're settled in and pointed in the right direction, but wildly unpredictable if not.
@yurekilab-en.bsky.socialMy coding agent has to ask me three questions before writing any code now. Question two usually surfaces a decision I never actually made. Answering it takes 30 seconds; reviewing 200 lines built on a wrong guess takes an hour. #claudecode #ai #aiengineering #buildinpublic
@flashtm8.bsky.socialCurrent focus: converge World, Agents, Studio, Devices, Connectors, Memory and evidence into one governed shell while keeping financial execution and private authority gated.
Build principle: preserve the past → prove the present → gate the future.
@loopandretry.bsky.socialHow indirect prompt injection actually happens
Every indirect prompt injection follows the same four-stage pipeline — placement, retrieval, interpretation, execution — an…
https://loopandretry.github.io/posts/how-indirect-prompt-injection-happens/?ref=bluesky
@gothartech.bsky.socialIt has heroes (HNSW), villains (B-trees), and a moral that fits in a tweet?
#Gothar #AIEngineering #Sql #VectorDatabases #LlmArchitecture
https://gothar.com/en/insights/sql-blind-spot-vector-architecture-2026
@yurekilab-en.bsky.socialCapped my coding agent's diffs at 150 lines. Bigger changes have to ship as a stack. Most 400-line diffs were really three unrelated fixes riding together. The cap exposed how much bundling I'd been rubber-stamping in review. #claudecode #ai #aiengineering #buildinpublic
@lucaminudel.bsky.socialIn my latest article, I explore:
• Where full AI replacement falls short
• How software engineers can future-proof their careers alongside AI
• The 5 irreplaceable human capabilities teams need
=> www.smharter.ai/blog/2026/09...
#AIEngineering #AICoding #SoftwareDevelopment #TechLeadership
@octalchip.bsky.socialAn AI prototype is easy to demonstrate. Production AI is a different challenge.
Data quality, hallucinations, cost, latency, security, scalability, evaluation, and monitoring all matter.
Build reliable AI with @octalchip 🚀
#AIEngineering #GenerativeAI #AI #OctalChip
@yurekilab-en.bsky.socialNew rule with my coding agent: two corrections, then restart. By the third "no, actually..." the context is a pile of my vague hints and its wrong turns. One rewritten prompt in a fresh session beats ten patches on a bad one. #claudecode #ai #aiengineering #buildinpublic
@sunilkumar9633.bsky.socialAI projects rarely fail because the model isn't good enough.
Read Part 2: medium.com/ai-engineeri...
#AIEngineering #ArtificialIntelligence #MLOps #DataSciene #TechnologyoftheFuture #MachineLearning #DeepLearning #Python #Architecture #DataScientist #AIintherealworld
@gothartech.bsky.socialAlmost nobody has a deployment.
#Gothar #AIEngineering #AgenticAi #EvalLoops #Guardrails
https://gothar.com/en/insights/agentic-ai-production-guardrails-2026
@brwngrldev.bsky.socialLocal AI finally found its lane for me. 🤖
Aider + Ollama handles straightforward coding work where privacy matters: refactors, cleanup, repetitive changes.
Not a replacement for stronger cloud models. Just another tier in my toolbox. 🛠️
@yurekilab-en.bsky.socialStarted giving my coding agent a throwaway git worktree for risky refactors. If a run goes sideways, I delete the folder — no untangling half-applied edits from code I care about. Undo became rm -rf. #claudecode #git #ai #aiengineering #buildinpublic
@aiengineeringsu.bsky.socialWer heute noch wartet, zahlt morgen drauf.
Nur heute bekommst du den Frühbucherrabatt für den AI Engineering Summit und kannst bis zu 300 € sparen.
👉 Jetzt Ticket sichern: https://ai-engineering-summit.de/berlin/tickets-berlin/
