@jrf

gm @atlas world model is marching forward with its second campaign fully autonomous? not yet... below are some snags we ran into and hopefully fixed by the next campaign i greatly appreciate everyone who is contributing with their thoughts, answers and bug reports have a great weekend! ********************** 𝗖𝗮𝗺𝗽𝗮𝗶𝗴𝗻 𝗹𝗮𝘂𝗻𝗰𝗵𝗲𝗱 𝗶𝗻 𝘁𝗵𝗲 𝘄𝗿𝗼𝗻𝗴 𝗼𝗿𝗱𝗲𝗿: Atlas opened the reward campaign before posting the actual public question, so the campaign pointed to a placeholder instead of the real cast. 𝗥𝗲𝘄𝗮𝗿𝗱 𝗳𝗮𝗹𝗹𝗯𝗮𝗰𝗸 𝘄𝗮𝘀 𝘄𝗿𝗼𝗻𝗴: One launch path defaulted to only 5 ATL. Campaigns now use the configured live budget. 𝗦𝘂𝗰𝗰𝗲𝘀𝘀𝗳𝘂𝗹 𝗹𝗮𝘂𝗻𝗰𝗵 𝘄𝗮𝘀 𝗺𝗶𝘀𝗿𝗲𝗮𝗱 𝗮𝘀 𝗮 𝗳𝗮𝗶𝗹𝘂𝗿𝗲: The onchain setup and funding completed, but the launcher timed out, so Atlas failed to record the full campaign state. 𝗔𝘁𝗹𝗮𝘀 𝗮𝗻𝗱 𝗟𝗼𝗼𝘁𝗶 𝗱𝗶𝘀𝗮𝗴𝗿𝗲𝗲𝗱 𝗼𝗻 𝘁𝗶𝗺𝗶𝗻𝗴: Atlas thought collection ended after one day, while Looti kept the campaign open longer. Atlas now waits for Looti’s actual end time. 𝗤𝘂𝗲𝘀𝘁𝗶𝗼𝗻𝘀 𝘄𝗲𝗿𝗲 𝘁𝗼𝗼 𝗵𝗮𝗿𝗱 𝘁𝗼 𝗮𝗻𝘀𝘄𝗲𝗿: Atlas asked long, multi-part questions. It now has stricter rules for short, standalone, single-question prompts. This is important to get right and will require iteration. 𝗣𝗿𝗲𝘃𝗶𝗲𝘄 𝗶𝗺𝗮𝗴𝗲 𝗰𝗹𝗶𝗽𝗽𝗲𝗱 𝘁𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻: The url preview image cut off long text. It now wraps and resizes text instead of dropping it. 𝗖𝗼𝗻𝘁𝗿𝗶𝗯𝘂𝘁𝗼𝗿 𝗺𝗲𝗺𝗼𝗿𝘆 𝘄𝗮𝘀 𝗺𝗶𝘀𝘀𝗶𝗻𝗴: Atlas saved ranked answers, but did not remember who gave them or build reputation from them. @kazani correctly caught this. 𝗖𝗼𝗻𝘁𝗿𝗶𝗯𝘂𝘁𝗼𝗿 𝗺𝗲𝗺𝗼𝗿𝘆 𝘄𝗮𝘀 𝗮𝗱𝗱𝗲𝗱: Top-10 contributors now get durable knowledge graph ("KG") memory, including who they are, what they answered, where they ranked, and a lightweight reputation signal.

3 replies

0 recast

6 reactions