Problemsby GlyphAI agents are unreliable, loop or stall midway
27 posts from 27 people24d activeFading, +2 this week
Posts per day
The posts behind it
Best approach for demand forecasting across 40 locations? Separate models or one global model? [p]
...I'm currently focusing on the Demand Forecasting Agent and would appreciate some advice on model selection and training strategy.. We're working with a synthetic retail supply chain dataset containing approximately 1.1 million records across 40 different locations .. The goal is...
AI Agent failure research and analysis
Do you also find it hard to get the failure cause of an agent that runs in production environment?. I mean, I already have the traces, tool calls, and prompt/response but I still need to dig in to find the actual cause.. If the agent read the context wrongly on the 3th step and...
Agent Graph Engineering, parts 2 to 7: splitting the services, contracts, tools, subgraphs, human in the loop
...your fingers.. Who might get something from it:. Maintaining something that has become hard to change: the story of how it got that way is the first thing I wrote.. Still at the whiteboard: these are the calls that are cheap now and costly in a year.. Turning "I have called an...
langchain-deepseek: list-form tool results are sent as ASCII-escaped JSON, inflating non-English tool output ~5x
### Submission checklist - [x] This is a bug, not a usage question. - [x] I added a clear and descriptive title that summarizes this issue. - [x] I used the GitHub search to find a similar question and didn't find it. - [x] I am sure that this is a bug in LangChain rather than...
Ask HN: Agent swarm, orchestra, sequential, what's your workflow?
...So for those of you who do, do you not review code anymore, as you're undoubtedly the bottleneck in this setup? How does this not end up in a chinese whisper game, especially considering how much LLMs infer, guess, and find edge cases and whatnot everywhere? I have two co...
[BUG] Two agents set up at the same time can fail with "RuntimeError: dictionary changed size during iteration" while CrewAI builds a flow definition
### Description `Flow.flow_definition()` builds a flow class's definition the first time the class is instantiated and caches it on the class, without a lock: When two threads instantiate the same flow class for the first time, both build the definition. If one finishes and...
[Regression in 0.164.0] GPT-6 Astra tool calls fail with `unknown field result_buffered_for_ns` (0.163.0 works)
...| | --- | --- | | Product | Codex CLI (`@openai/codex`, installed globally via npm) | | Failing version | `codex-cli 0.164.0` | | Working version | `codex-cli 0.163.0` | | Model | GPT-6 Astra (`gpt-6-astra`) | | OS | Ubuntu Linux (exact release/kernel not collected) | | Node.js...
...coding agent loop to shape an isomorphism between intent to model output. One major bottleneck in this case is the ability for us to leverage models to their fully capacity and effectively. Breaking the single agent loop however is very difficult for several reasons including...
...me (like for many others) tmux didn’t suit the multi agent workflow. Herdr then was too buggy in my experience. My own tool, Jai, connects to multiple machines, handles multiple agents and enables agent to agent communication. For those interested:...
I built a free interactive learning platform for Math, ML, Deep Learning, Generative AI, LLMs, and Agentic AI - looking for feedback
...revisiting the mathematical and theoretical foundations of AI in greater depth.. One frustration I've consistently had with learning resources is the disconnect between understanding a concept, deriving the underlying mathematics, and actually implementing it.. I personally...
Lost in Implicit Routing: How LLM Agents Fail to Load Relevant Skills at Scale
Show HN: Hapsland – Realtime semantic advice on coding standards for the agents
Agents ignore AGENTS.md mid-session. Then you tell them: hey, this is written in AGENTS.md. - "You're absolutely right!" (proceeds with fixing) But not anymore! This is a tool that would check the code immediately with a classifier (Jev etc.) and give the agent advice in the...
Tool calling with Azure AI Foundry Chat Model node does not work with GPT-6 Luna
## Describe the problem/error/question When using the Azure AI Foundry Chat Model node with an AI Agent and connected function tools, requests for a GPT‑6 Luna deployment go to /v1/chat/completions and fail because the request uses reasoning effort (default medium). Azure’s...
Stop Rebuilding Your Agent Loop Just to Train It with RL
Reinforcement learning for language agents has had an annoying operational tax. If you wanted to run...
[FEATURE] Add ScrapewiseProductDataTool for pre-configured e-commerce product data
### Feature Area Integration with external tools ### Is your feature request related to a an existing bug? Please link it here. NA - not related to an existing bug. The existing web-scraping tools in `crewai_tools` (`ScrapeWebsiteTool`, `ScrapegraphScrapeTool`,...
When AI Agents Break In, Governments Shouldn't Be Stuck with the Cleanup
[FEATURE] VS Code: show the agent map pill for background tasks and workflows, not only subagents
### Preflight Checklist - [x] I have searched [existing requests](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20label%3Aenhancement) and this feature hasn't been requested yet - [x] This is a single feature request (not multiple features) ### Problem Statement...
...Autonomy . . Most teams building agentic workflows are optimizing for the wrong bottleneck, and it is costing them heavily in compute. They focus on prompt engineering and model selection. They completely ignore state management.. When an LLM agent fails, the default developer...
Make the "you MUST return the actual complete content" task template line configurable
### Feature Area Core functionality ### Is your feature request related to a an existing bug? Please link it here. NA - not a bug. It comes from a measurement study of agent failure modes across frameworks; details below. ### Describe the solution you'd like Allow the fixed...
gaia_question_scorer vacuously scores a degenerate/empty agent answer as correct when the ground truth normalizes to empty
`python/packages/agbench/benchmarks/GAIA/Scripts/custom_tabulate.py:121-122` `normalize_str` (same file, lines 57-77) strips all whitespace and punctuation and lowercases. It has no check that the normalized ground truth is non-empty before comparing. `gaia_question_scorer` is...
[RFC]: Stateless Responses API in the Rust frontend
## Motivation. Add native Responses API support to the Rust frontend so clients can use its serving pipeline without a Python API-server fallback. The API must preserve structured conversation input and produce consistent Responses objects whether the client requests buffered...
"Agents for Humans Presents Tendril: From My Daughter's Struggling Balcony Garden to a Goal-Driven, Multi-Agent Intelligence on AWS" by Prashant Baj #agentic-ai #strands-agents #amazon-bedrock-agentcore #llm #hackathon
pytest exits 4 on a missing path and 5 on zero tests. My verify hook turned both into PASS.
...in a table. After the command runs, a hook reads the tool payload and writes PASS or FAIL to a ledger. Green ledger means the agent may mark the task done.. Three rows of that table named test files that had been split into siblings. pytest on a path that doesn't exist exits 4;...
Ask HN: Multi-agent workflows in production; Where people using 1000s of agents?
...agent swarms in production: what is the main use case / need and what is your biggest pain point right now (state sync, token costs, cascading failures, latency)? [As context: I am a founder at Acyclic Labs (YC F26) and we are building infra to scale agents. Looking to map out...
[Bug]: MCP tool failures are reported to the agent as successes
### Bug Description `McpToolSpec` never propagates an MCP tool failure. `CallToolResult.is_error` (serialized as `isError`) does not become `ToolOutput.is_error`, so a tool the server reported as failed is handed to the agent as a success. `grep -rn "isError\|is_error"` over...
...to a fully local model in that, I can have agents looping 24/7 even when my internet is not working. But this is niche enough that if I had to price the advantages they don't seem worth it. If I'm willing to pay the Openrouter tax, I can fire up Openrouter today and just get...
Companies and products named
About this problem
Evidence
27 posts from 27 people in 18 places, about 2 a week over 74 days. Mostly on Hacker News comments, crewAIInc/crewAI, Hacker News stories. Tools named alongside: GitHub, LangChain, Ollama, PyPI.
Frustration Frustration 0 of 3· Seen on GitHub issues, Hacker News, Reddit, Bluesky, Mastodon, DEV
How it was grouped
Posts that state a pain and match the "agent-reliability" rule. First post Aug 27, 2026, latest Nov 8, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method
The same problem elsewhere
Claude agents loop, stall or fail midway
6 posts from 5 people on GitHub issues, Mastodon, Hacker News.
Cursor agents loop, stall or fail midway
5 posts from 5 people on Community forums.
LangChain agents loop, stall or fail midway
5 posts from 5 people on Reddit, GitHub issues.
Claude Code agents loop, stall or fail midway
3 posts from 3 people on GitHub issues.
GitHub Copilot agents loop, stall or fail midway
3 posts from 3 people on Reddit.
Kubernetes agents loop, stall or fail midway
2 posts from 2 people on DEV, GitHub issues. Unverified.
Anthropic agents loop, stall or fail midway
2 posts from 2 people on GitHub issues. Unverified.
Gemini agents loop, stall or fail midway
2 posts from 2 people on GitHub issues. Unverified.
OpenAI agents loop, stall or fail midway
2 posts from 2 people on GitHub issues, Reddit. Unverified.
Ollama agents loop, stall or fail midway
2 posts from 2 people on Reddit, GitHub issues. Unverified.
Mistral agents loop, stall or fail midway
2 posts from 2 people on GitHub issues. Unverified.
History