tweetindex

安叫兽|Bird🕊️ 🔶 BNB

@ajs6888 · 撸毛卡网: · joined 24 Apr 2022

AI中转站:https://t.co/TpSOi0jwcp|其他所返佣https://t.co/1YRGFH5Rwv 币安永久返佣:https://t.co/BVCUkKvBQM|欧意:https://t.co/xlGiGnQFNU

19 886Followers
3 114Following
116 045Posts total
1.7KViews on collected posts

Against accounts of the same size

5 posts from the last 90 days, next to the 10K–100K follower range. reaches fewer people than peers, but engages them much harder.

Median views311this account2 259median for 10K–100K
Reach, %1.56%this account6.80%median for 10K–100K
Engagement, %1.96%this account1.52%median for 10K–100K
MetricThis accountMedian for 10K–100KRatio
Median views per post3112 2590.14×
Reach (views ÷ followers)1.56%6.80%0.23×
Engagement rate1.96%1.52%1.29×

Others in this range →   Compare with another account →   How these benchmarks are built →

Growth & engagement

How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.

Views per post

3112 Sep
459
3663 Sep
299
222

Last 5 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.64%2 Sep
1.96%
3.28%3 Sep
0.67%
4.05%

Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.

What the audience does

Likes50.0%17 in total
Replies50.0%17 in total

Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.

The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.

Latest posts

260M 参数的 NeoMME 一页图像索引从约 1.5MB 压到 6kB 不是把模型做大 是把视觉检索这件事做得更像工程 Hugging Face 这篇写得比较实 2048×2048 输入 在 L40S 上每秒编码约 51 页 还保留了 95% 以上的 nDCG@10 我更在意这个取舍 RAG 场景里索引成本经常比模型参数更先撞墙 但别把 260M 222 views · 3 likes · 0 reposts · 6 replies 03 Sep 2026 换一个 coding agent 上周的排查记录就没了 Hugging Face 今天发了 funes 把 Claude Code Codex pi 和 Hermes 的会话记录做成可检索记忆 不是再开一个聊天窗口 它会从旧 session 里找出 当时为什么放弃某条方案 哪个错误已经踩过 再把原始出处带回来 我觉得这比“让 Agent 记住我的偏好”实在 299 views · 1 likes · 0 reposts · 1 replies 03 Sep 2026 Reef 这个仓库有点像给 Agent 加了一本会更新的操作手册 但它先做的不是让模型变聪明 它把一次请求 反馈 训练更新 版本提交拆成四步 模型权重和 harness 都能走这条链 我比较在意最后那个 commit 没有评估和版本记录的自我改进 多数时候只是 Agent 自己说自己变好了 https://t.co/8c2ntKNYtz 366 views · 6 likes · 0 reposts · 6 replies 03 Sep 2026 Databricks 7个 MCP 小 bug 一年烧掉约 50 万美元 token 还搭进去 1.2 万工程师小时 这是 Databricks 自己拆的一次线上账 他们用 trace 把失败调用排出来 一个小时就找到了最该修的那批 不是先换更便宜的模型 是先别让 Agent 对着坏工具反复重试 我觉得很多团队的成本账都算反了 天天盯模型单价 459 views · 5 likes · 0 reposts · 4 replies 02 Sep 2026 Aura 在 HN 上只有 15 分 2 条评论 但我更在意它的 README 写了什么 它不是让 Agent 多开几个终端 是把支付故障丢给 3 个专门 Agent 去对 traces logs 和 Kubernetes 历史 最后定位到 productcatalogservice 1.13.2 的 N+1 回归 还给出回滚到 1.13.1 的建议和恢复检查 这种 SRE Agent 我才会认真看 311 views · 2 likes · 0 reposts · 0 replies 02 Sep 2026

Similar accounts