Pithy Cyborg | AI News Made Simple
Pithy Cyborg | AI News Made Simple
An AI coding assistant does not remember that it fixed a bug. Each new request sees the current state of the code and the current conversation context. If the fix was applied to one file but the bug's …
Continue Reading about Why Does My AI Coding Assistant Keep Reintroducing the Bug It Just Fixed? →
Yes. A retry wrapper on a non-idempotent tool call retries an operation that may have already succeeded. The model called the tool, the tool executed, the response was lost due to a timeout or network …
Continue Reading about Can a Retry Wrapper Make LLM Tool Calls Less Reliable Instead of More? →
Streaming sends tokens as they are generated rather than waiting for the complete response. When the complete response is valid JSON, streaming appears to work. The problem appears at the consumer: …
Continue Reading about Why Does Streaming JSON From an LLM Break Even When the Output Looks Valid? →
Yes. A PDF passed to an AI workflow is treated as trusted content by default. If that PDF contains text formatted as instructions rather than content, the model reading it may follow those …
Continue Reading about Can Prompt Injection Hide Inside a PDF and Poison an AI Workflow? →
Removing internet access from an AI chatbot closes one data exfiltration channel. It does not close the others. Logs, shared context windows, prompt injection via uploaded files, insecure API key …
Local LLM inference is sequential by default. When one user sends a request, the model generates tokens one at a time until the response is complete. When three users send requests simultaneously, …
Continue Reading about Why Does a Self-Hosted Model Feel Fast for One User but Collapse for Three? →
