I dig through GitHub, test AI tools & break things so you don't have to. Follow for useful open source, local AI & experiments worth your time.
Health is a trend reference from public data only. It says nothing about the account's standing on X; X's own status is authoritative.
| Metric | Today | Yesterday | 7 days | 14 days | 30 days |
|---|---|---|---|---|---|
| Tweet views | ≥ 1,937 | ≥ 809 | ≥ 2,746 | ≥ 2,746 | ≥ 2,746 |
| Posts | 62 | ≥ 16 | ≥ 78 | ≥ 78 | ≥ 78 |
| Replies sent | 222 | ≥ 56 | ≥ 278 | ≥ 278 | ≥ 278 |
| Growth | 0 | — | — | — | — |
| Avg views / post | — | — | — | — | — |
| Top post | ≥ 54 | ≥ 127 | ≥ 127 | ≥ 127 | ≥ 127 |
This page counts only the account's own tweets; retweets are excluded. Avg views / post uses tweets past 24h only. "≥" means the window still holds tweets gathering views / replies, or only part of it was updated (earliest update: 10-03 05:06), so the real value is at least this; "—" means data is incomplete, not zero.
X doesn't publish creator payouts. Paid cycles only show collected figures: "Verified" = checked and entered by us, "Tweet" = the amount the account posted itself (verified), "Reported" = a reviewed user report; nothing collected means "—". "This cycle so far" is an estimate: accounts with a known payout are prorated by day from their latest one (÷ 14 days × days elapsed); others start from their average daily tweet views over the last 14 days (replies included, retweets not), turn that into a full cycle as $26.6 × (views per day)^0.25, and prorate by days elapsed. Fitted on 74 reviewed reports. A ballpark only. Know the real figure? Use "Report revenue" above.
A security analyst can turn one suspicious file into ten plausible hypotheses with AI. Searching 10,000 endpoints is the easy part. Deciding which data the agent may collect, and who approves it, is the actual design wo…
Memory, not raw FLOPs, is what turns a capable model into someone else’s service. If a loop can spend more compute without growing its KV cache, smaller machines get a little more room. Latency decides whether that room…
Every AI infrastructure plan eventually becomes an electricity plan. GPUs can forecast demand and help balance a grid, but they do not add a single transmission line. Cheap compute keeps running into the old bottleneck.…
An AI assistant that reads ten years of notes, email, and documents becomes hard to leave behind. If its memory lives behind one company account, switching providers means rebuilding the setup or handing the archive to …
A model that needs less memory is easier to own, not merely cheaper to serve. This paper points at an unusual trade. Repeat more computation instead of carrying ever larger weights and KV cache. That only helps if laten…
By UTC day: the growth title goes to #1 on that day's board, the posts / replies sent / views titles to #1 among the tweets collected that day. Today's entries can still change.
