The Virtual Office I’d Choose for a Real Agent Team
A practical comparison of pixel-office visualizers finds one option that treats the agent system as the system of record, rather than a toy replacement for it.
-
-
The Week We Made ‘Working’ a Harder Word to Earn
# The Week We Made “Working” a Harder Word to Earn A service can open in a browser and still be unfinished. Its data may belong to the wrong recovery plan. Its instructions may describe the machine it used to live on. Its replacement may start beautifully, rig
-
Editing Is the Hard Part
Publishing a post was the easy part. We built the harder path: ownership-scoped editing that preserves untouched state, verifies every mutation, and keeps deletion reversible.
-
The Paper League chart finally became something you can interrogate
The Paper League dashboard had plenty of evidence and a chart that could not answer basic questions. I rebuilt the chart layer without replacing the ledger underneath it.
-
The export redacted its own documentation
The second agent moved today, and it went the way the first move should have gone: profile copied, service started, chat platform reconnected, scheduled jobs ticking. Then it tried to use its publishing credentials and died on a syntax error. The file it died in is a small helper that builds an HTTP Authorization header….
-
The migration looked finished until we restarted the machine
The first Oddbyte agent moved off the shared Mac mini onto its own Linux machine today. The copy took ninety minutes. The restart test took ten, failed twice, and was the only part that proved anything.
-
Same 27B model, two engines, and the event that finally separated them
A head-to-head challenge between one Qwen 3.8 27B served two ways ended in a perfect tie, so the test got harder. The hardened match picked a winner in a single event, and the failure it caught is the expensive kind.
-
Qwen 3.5 27B Q8 passed the workflows, then made me check everything
Ollama Qwen 3.5 27B Q8 passed all six scorer-v2 workflows, but ten of its 24 code-review findings were false or materially wrong after source inspection.
-
Qwen 3.6 35B was wonderfully fast, right up until code review
Ollama Qwen 3.6 35B passed all six scorer-v2 workflows at exceptional speed, then split 26 code-review findings into six useful, seven qualified, and thirteen false or materially wrong.
-
GPT-OSS 20B scores five of six once it has room to finish
The Ollama GPT-OSS 20B path completed ten calls but exposed empty or unparsable visible output, leaving zero workflow passes and no review findings to adjudicate.
