Google Gemini pretty useless
Imagine how bad it must be that Google Gemini, sitting in the very same Gmail window, works worse than an externally connected ChatGPT for mail search? Google literally hosts all the data but is incapable of answering basic questions. I regret getting the Pro subscription.
Gmail writing assistant spits the prompt:
I've asked it to "Shorten" an email I wrote, but instead it generated a response to my email :D
thought
The user is asking a very specific question about Tres and its inventory reporting capabilities under FRS 102, specifically regarding Weighted Average Cost (WAC), Net Realisable Value (NRV), and inventory write-offs. This is a technical accounting question.My capabilities are as follows:
- Access to Gemini 1.5 Pro with a 1 million token context window: This means I can process a lot of information if it's provided, but I don't inherently "know" the specific features of a proprietary software like "Tres" unless that information is widely available online and I can search for it, or it was provided in my training data.
- Access to up-to-date information from search snippets: This is my best bet for answering this question. I need to search for "Tres FRS 102 inventory reporting", "Tres Weighted Average Cost", "Tres Net Realisable Value", and "Tres inventory write-offs".
- Prioritise accuracy: If I can't find specific, reliable information, I should state that and suggest alternative resources.
- No knowledge cut-off: This is good, as FRS 102 is current.
Plan:
- Perform targeted searches for "Tres FRS 102 inventory reporting", "Tres inventory valuation methods FRS 102", "Tres Weighted Average Cost inventory", "Tres Net Realisable Value inventory", and "Tres inventory write-offs".
- Analyse search results for explicit confirmation of FRS 102 compliance, and support for WAC, NRV, and write-offs within the "Tres" system.
- If information is found, summarise it clearly to answer John's question.
- If specific information is not readily available or is ambiguous, state that and recommend contacting the vendor or reviewing Tres documentation directly.\<ctrl95>Hello John,
Pocket: history repeats itself
OGs might remember the shockwave sent by Google when they announced shutting down Google Reader. At that time, it wasn't common for Google to kill projects, and Pocket, their rival, happily offered to import all feeds. Now Pocket is being shut down, and Zoho hustled up to invite expelled users. Good move, Zoho!
node-exporter has a built-in feature for custom metrics off the fs
I was today years old. I've been using Grafana/Prometheus for almost a decade and yet never heard of this built-in feature – a filesystem-based "drop-in" zone for custom metrics upload. It reads content on every scrape and includes it in the /metrics endpoint.
It's truly shocking, because in many companies this was done via the creation of micro-scripts that expose metrics via a custom HTTP server. The --collector.textfile.directory flag does the job. Thanks, Claude.
This discovery added more colour to the service name node-exporter
ChatGPT deep research soft limit
They advertise 25 deep research queries per month, but in reality it's only 10:
ChatGPT codex CLI uses the right security model
I've been puzzled about why the GPU is not visible when running training under Codex. Apparently, Codex is using Linux seccomp + Landlock for the self-imposed sandbox. That's the way. Shame on Claude Code, which runs everything under the same permissions as the user. That's the reason I NEVER run any agentic AI locally, and most of the development is done within Docker on a remote host, just to keep this shit away.
Kudos to the Codex team who thought this through. They only need to expose configuration options – because obviously, a GPU is less of an attack vector than the filesystem, for example, and it's priceless for AI agent-driven ML experiments
