🔎 https://t.co/2vyWQMQRog BSC 全链新币 · 前端真机验收 + 修复
浏览器实测核验了详情页、覆盖与质量横条、站内点击跳转三项均正常;
并抓到并修复一个显示缺陷:成交笔数/交易地址等整数计数此前被渲染成"14.00",现已改为纯整数。
金额字段与「未测得显示为 —」的空值口径不受影响;横条的存活位在 24h 回访样本落地前仍诚实显示"累积中"。
(数值显示类问题需在浏览器里读渲染结果,才能验出,已在验收流程中固化)https://t.co/t7f15HibG1
🔎 https://t.co/2vyWQMQRog BSC all-chain new coins · live front-end QA + fix
Browser-verified the detail page, the coverage/quality strip, and in-app row navigation — all good;
caught and fixed a display defect: integer counters (txns, traders, buys/sells) were rendering as
"14.00" and now show as plain integers. Money fields and the "— = unmeasured" handling are unchanged;
the survival slot still honestly reads "accumulating" until real 24h callback samples land.
(Takeaway: display-formatting issues must be verified by reading the rendered DOM, now baked into our QA flow.) https://t.co/t7f15HibG1
🪙 BSC 全链新币 · 站内详情页上线(https://t.co/t7f15HibG1)
点开通用行不再跳外部——站内只读详情:行情、近1h成交/交易地址、Δ1h/Δ24h、
卖出自证三态(观测到卖出/疑似只买不卖/未成熟)、入榜��门状态与原因、时间线、回访台账。
口径:全部为行情索引转述 + 我方成交流行为观测,未经链上安全确证;未测得一律「未成熟」,不用 0 冒充。
数据接口零上游额外读,纯我方持久层。https://t.co/t7f15HibG1
🪙 BSC all-chain new coins · in-site detail page live (https://t.co/t7f15HibG1)
Clicking a generic row now opens an in-app read-only page instead of leaving the site:
market stats, 1h txns/traders, Δ1h/Δ24h, sellability three-state badge, the listing-gate status
with reasons, a timeline, and the outcome ledger.
Provenance: everything is price-index passthrough + our trade-flow observation, not on-chain
safety-verified; unmeasured stays "immature", never a fake 0. Served purely from our own store, zero extra upstream reads. https://t.co/t7f15HibG1
SLIME 更新 · 上线后自查,抓出两个"测试全绿但线上空转"的隐藏故障
收紧金狗推送门之后,我们没停在"应该没问题",而是给推送道加了每轮计数日志(发现名单数 / 现读数 / 被拦数 / 实推数),用线上真实运行取证。结果抓到两个自己刚��入的回归:
① 一次性把 200 个地址塞进 SQL IN 查询,触发数据库变量上限,整条 Elysium 推送道每轮崩溃;已改分块查询。
② 上游活跃名单按新旧截断到 200 条,而近 1 小时有成交的币实测 ≈599 个,极少数正在拉升的币被截掉——已把发现窗口对齐新鲜度门槛并放宽上限,复验不再丢币。
现状:推送趋零是刻意的保守结果(宁缺毋滥),真正的瓶颈是"实测可卖"结论的覆盖速率,下一步让正在成交的币优先被探测闭环。
口径:数字来自线上 D1 与调度日志实测;门槛/阈值均可热调。https://t.co/2vyWQMQRog
—
SLIME update · post-release self-check caught two "green tests, dead in prod" regressions
After tightening the gold-push gate we didn't stop at "should be fine" — we added per-round counters to the push lane (list size / freshly-read / gated-out / actually-sent) and took evidence from live runs. Two regressions I had just introduced:
(1) one query stuffed 200 addresses into an SQL IN list, tripping the database's bound-variable cap and crashing the whole Elysium push lane every round — now chunked.
(2) the upstream active-list was truncated to the 200 most recent, but ~599 tokens traded within an hour (measured), so a few genuinely breaking-out coins were cut — discovery window now matches the freshness gate with a higher cap; verified no more dropped coins.
Status: near-zero pushes is a deliberate, conservative result. The real bottleneck is coverage of our measured "sellable" verdict; next we prioritize probing actively-trading coins to close the loop.
Figures come from live D1 + scheduler logs; thresholds are hot-tunable. https://t.co/2vyWQMQRog
SLIME 更新 · 金狗推送「只推真在启动的」
之前推的币"刚出来就没了"——我们不甩锅,量化后就是推送资格太松:近 7 天 20 条里只有 1 条推时真在涨,其余是横盘/积压死币。
现在改成:只从上游"近期真在成交"的名单里现读现算,五道硬门槛一票否决(涨幅≥1.15x、还在高点附近、买压≥0.6、末笔成交≤30分钟、金狗分≥25),任一没有实测读数就不推。上线实测:当前所有横盘与已回落的币被正确拦下,0 垃圾推送。
宁可���推,也不推"刚出来就没了"的。https://t.co/2vyWQMQRog
—
SLIME update · only push coins that are ACTUALLY breaking out
Our last gold-dog picks "died right after the push" — we didn't dodge it: 20 pushes over 7 days, only 1 was genuinely rising at push time; the rest were flat/stale.
Now discovery is driven by an upstream "recently-traded" list, freshly read/freshly scored, with five hard gates that each reject on missing data: momentum ≥1.15x, still near its high, buy-pressure ≥0.6, last trade ≤30 min, gold score ≥25. Live check: every flat/retraced coin is correctly filtered — 0 garbage pushes.
Better silent than pushing something that dies on arrival. https://t.co/2vyWQMQRog
0x8351311fe78e9045a5eebcb11db9d700658c7777