#!/usr/bin/env python3 """Deep analysis for xmon 2026-11-04. Sections: 1. Buckets (handle-group first, is_ai flag) + disjoint sum check 2. Raw core ranking + organic (like-rate >= 0.15%) ranking 3. 24h deltas vs 2026-11-03 on shared tweet ids 4. Per-handle freshness (max createdAt, in-window counts) 5. Originals-only dormancy 6. Offer / arbitrage grep over in-window originals 7. Key chronologies """ import json, re, os from datetime import datetime, timezone, timedelta RAW = "/home/kiro/trend-engine/trends/x/raw" TODAY, PREV = "2026-11-04", "2026-11-03" d = json.load(open(f"{RAW}/scan_results_{TODAY}_FULL.json")) p = json.load(open(f"{RAW}/scan_results_{PREV}_FULL.json")) raw_all = d["raw"] NOW = datetime.now(timezone.utc) CUT = NOW.replace(hour=0, minute=0, second=0, microsecond=0) - timedelta(days=3) IRRELEVANT = {"AIBuzzNews","AIGuideHQ","AITECHio","AIFrontliner","TheAIColony","AITOPIAai", "GoogleLabs","GoogleAI","DonutAI","cladlabs","FaithJaiye2187","KammelaBala","MyStageAI", "AlexStudio44","AI_with_jasmin","Ai_Vaidehi","ChrisLaubAI","CINEMA505","OpenAI","AnthropicAI", "AIScottAdams","AISuperDomain","AI_Jasonyu","AIByLeo","Soaima_Ai","different_ai","via_ai_", "Teamily_AI","AIsaOneHQ","AiWithTariq","numerai"} FILM = {"DiscussingFilm","nofilmschool"} NON_CORE = {"interesting_aIl","hubermanlab","NVIDIAAI"} def pd(s): try: return datetime.strptime(s, "%a %b %d %H:%M:%S %z %Y") except Exception: return None def score(t): return ((t.get("likeCount") or 0) + 2*(t.get("retweetCount") or 0) + 3*(t.get("replyCount") or 0) + (t.get("viewCount") or 0)/100) def qual(t): return ((t.get("likeCount") or 0) >= 50 or (t.get("retweetCount") or 0) >= 10 or (t.get("viewCount") or 0) >= 1000) def is_orig(t): return not (t.get("retweeted_tweet") or t.get("type") == "retweet" or (t.get("text") or "").startswith("RT @")) def mediatype(t): m = t.get("media") if isinstance(m, list) and m: return m[0].get("type", "?") return "-" # NOTE: leading AND trailing \b are both required. A leading-only \b matched # "ai" inside "aired" (re.I) and produced a FALSE AI-relevant positive in the # film bucket on 2026-11-04. Fixed here. STRONG = re.compile(r"\b(AI|A\.I\.|artificial intelligence|generative|LLM|GPT|diffusion|" r"machine learning|neural|prompts?|models? generat\w*)\b", re.I) # ---------- 1. BUCKETS ---------- buckets = {"core":[], "film":[], "irrelevant":[], "labs":[], "offtopic":[], "health":[], "new":[]} rows = [] for h, tw in raw_all.items(): for t in tw: dt = pd(t.get("createdAt","")) if not dt or dt < CUT or not is_orig(t) or not qual(t): continue if h in ("OpenAI","AnthropicAI","NVIDIAAI"): g = "labs" elif h == "interesting_aIl": g = "offtopic" elif h == "hubermanlab": g = "health" elif h in FILM: g = "film" elif h in IRRELEVANT: g = "irrelevant" else: g = "core" r = {"h":h,"g":g,"dt":dt,"score":score(t),"likes":t.get("likeCount") or 0, "rt":t.get("retweetCount") or 0,"replies":t.get("replyCount") or 0, "views":t.get("viewCount") or 0,"text":(t.get("text") or "").replace("\n"," ").strip(), "url":f"https://x.com/{h}/status/{t.get('id')}","id":str(t.get("id")), "media":mediatype(t), "is_ai":bool(STRONG.search((t.get("text") or ""))), "quoted":bool(t.get("quoted_tweet"))} buckets[g].append(r); rows.append(r) print("===== 1. BUCKETS (disjoint) =====") tot = 0 for g,v in buckets.items(): ai = sum(1 for x in v if x["is_ai"]) print(f" {g:12s} {len(v):4d} AI-relevant: {ai}") tot += len(v) print(f" SUM {tot} vs qualifying {len(d['qualifying'])} ->", "OK" if tot == len(d["qualifying"]) else "MISMATCH") # ---------- 2. RANKINGS ---------- core = buckets["core"] raw_rank = sorted(core, key=lambda x: -x["score"]) for x in raw_rank: x["lr"] = (x["likes"]/x["views"]*100) if x["views"] else 0 org = sorted([x for x in core if x["lr"] >= 0.15], key=lambda x: -x["score"]) print("\n===== 2a. CORE RAW TOP 12 (score) =====") for i,x in enumerate(raw_rank[:12],1): print(f"{i:2d}|@{x['h']}|{x['likes']}L|{x['rt']}RT|{x['replies']}R|{x['views']}V|" f"lr={x['lr']:.2f}%|{x['dt'].strftime('%m-%d %H:%M')}|scr={x['score']:.0f}|{x['media']}|{x['text'][:110]}") print("\n===== 2b. CORE ORGANIC TOP 12 (like rate >= 0.15%) =====") for i,x in enumerate(org[:12],1): print(f"{i:2d}|@{x['h']}|{x['likes']}L|{x['rt']}RT|{x['replies']}R|{x['views']}V|" f"lr={x['lr']:.2f}%|{x['dt'].strftime('%m-%d %H:%M')}|scr={x['score']:.0f}|{x['text'][:110]}") if len(org) > 12: print(f" cutoff scr={org[9]['score']:.0f} rank11 scr={org[10]['score']:.0f}") print("\n===== 2c. PAID-REACH WATCH (view-heavy, like rate < 0.15%) =====") for x in raw_rank[:14]: if x["lr"] < 0.15: print(f" @{x['h']}|{x['likes']}L|{x['views']}V|lr={x['lr']:.3f}%|{x['text'][:80]}") # ---------- 3. 24H DELTAS ---------- def idmap(js): m = {} for h, tw in js["raw"].items(): for t in tw: if is_orig(t) and t.get("id"): m[str(t["id"])] = (h, t) return m cur, prv = idmap(d), idmap(p) shared = set(cur) & set(prv) deltas = [] for i in shared: h,tc = cur[i]; _,tp = prv[i] dl = (tc.get("likeCount") or 0) - (tp.get("likeCount") or 0) dv = (tc.get("viewCount") or 0) - (tp.get("viewCount") or 0) if dv or dl: deltas.append((dl, dv, h, tc, i)) print(f"\n===== 3. 24H DELTAS ({len(shared)} shared ids, {len(deltas)} changed) =====") for dl,dv,h,t,i in sorted(deltas, key=lambda z:-z[1])[:15]: print(f" @{h}|L{dl:+d}|V{dv:+d}|now {t.get('likeCount')}L/{t.get('viewCount')}V|{t.get('text','')[:80]}") print(" -- biggest LIKE gains --") for dl,dv,h,t,i in sorted(deltas, key=lambda z:-z[0])[:8]: if dl > 0: print(f" @{h}|L{dl:+d}|V{dv:+d}|{t.get('text','')[:80]}") print(" -- NEGATIVE like deltas --", sum(1 for z in deltas if z[0] < 0)) for dl,dv,h,t,i in sorted(deltas, key=lambda z:z[0])[:5]: if dl < 0: print(f" @{h}|L{dl:+d}|V{dv:+d}|{t.get('text','')[:80]}") # ---------- 4. FRESHNESS ---------- print("\n===== 4. FRESHNESS =====") fr = [] for h, tw in raw_all.items(): dts = [pd(t.get("createdAt","")) for t in tw] dts = [x for x in dts if x] if not dts: continue mx = max(dts); hrs = (NOW-mx).total_seconds()/3600 inw = sum(1 for t in tw if (pd(t.get("createdAt","")) or CUT-timedelta(days=1)) >= CUT and is_orig(t)) fr.append((hrs, h, mx, inw, len(tw))) fr.sort() print(" freshest:") for hrs,h,mx,inw,n in fr[:12]: print(f" @{h} {hrs:6.1f}h n={n} inwin={inw}") print(" key accounts:") for hrs,h,mx,inw,n in fr: if h in ("NVIDIAAI","venturetwins","EHuanglu","pika_labs","Kling_ai","runwayml","higgsfield", "higgsfield_ai","Flovaai","AIWarper","CharaspowerAI","c_valenzuelab","azed_ai", "heyglif","ElsaSofia__AI","ItsMaryamAi","AI__TSUBAKI","openart_ai"): print(f" @{h} {hrs:6.1f}h n={n} inwin={inw}") print(" stalest:") for hrs,h,mx,inw,n in fr[-10:]: print(f" @{h} {hrs:6.1f}h n={n} inwin={inw}") # ---------- 5. DORMANCY ---------- print("\n===== 5. DORMANCY (originals only) =====") watch = ["CuriousRefuge","karenxcheng","higgsfield","higgsfield_ai","WuxiaRocks","EinsiaAI", "pika_labs","AI__TSUBAKI","AIsaOneHQ","AIGuideHQ","AITOPIAai"] for h in watch: tw = raw_all.get(h, []) origs = [pd(t.get("createdAt","")) for t in tw if is_orig(t)] origs = [x for x in origs if x] if not origs: print(f" @{h}: NO originals in fetched set (n={len(tw)})"); continue mx = max(origs); hrs = (NOW-mx).total_seconds()/3600 inw = [x for x in origs if x >= CUT] best = 0 for t in tw: dt = pd(t.get("createdAt","")) if dt and dt >= CUT and is_orig(t): best = max(best, t.get("likeCount") or 0) verdict = "DORMANT" if not inw else f"ARTEFACT (best {best}L)" print(f" @{h}: newest orig {mx.strftime('%Y-%m-%d %H:%M')} ({hrs:.1f}h) inwin={len(inw)} -> {verdict}") # ---------- 6. OFFERS ---------- print("\n===== 6. OFFER / ARBITRAGE GREP =====") PATS = { "Flova launch special": r"(2 years of Pro|Buy 1 Year|price of 1|Launch Special)", "Flova credits/giveaway": r"(1,000 credits|credits each|Follow\+Repost|giveaway|winners)", "Pika H3": r"(H3 Flex|pika\.art|\$0\.01)", "Higgsfield offer": r"(unlimited|Genjutsu|Content Factory|Ads Studio|cashback)", "CapCut": r"(CapCut|capcutapp)", "DaVinci": r"(DaVinci|Resolve)", "Topaz": r"(Topaz|Starlight)", "Topview": r"(Topview|Topviewai)", "Chroma Awards": r"(Chroma|chromaawards)", "OpenArt": r"(openart|OpenArt)", "Seedance price": r"(\$0\.0[0-9]+|per second|/sec)", "Kling 4 / PixVerse": r"(Kling 4|PixVerse|Kling 4\.0)", "Runway dev": r"(dev\.runway|Runway API|Cursor)", "free credits/trial": r"(free generations|free credits|free trial|no credit|zero credit)", "discount/promo": r"(\d+% off|promo code|discount|coupon)", "Nebius": r"(Nebius|AI Studio|AI Builder)", "Magnific": r"(Magnific|Move app|Brand Kit)", "Dreamina": r"(Dreamina|DreaminaCPP)", "Seedream/Nano Banana": r"(Seedream|Nano Banana)", "Glif": r"(glif|Glif)", "HeyGen/avatar": r"(HeyGen|avatar|presenter)", } inw_all = [x for x in rows] for label, pat in PATS.items(): rx = re.compile(pat, re.I) hits = [x for x in inw_all if rx.search(x["text"])] if not hits: print(f" [{label}] 0 hits") continue print(f" [{label}] {len(hits)} hits") for x in sorted(hits, key=lambda z:-z["score"])[:4]: print(f" @{x['h']}|{x['likes']}L|{x['views']}V|{x['dt'].strftime('%m-%d %H:%M')}|{x['text'][:170]}") # ---------- 7. CHRONOLOGIES ---------- print("\n===== 7. CHRONOLOGIES =====") for h in ["EHuanglu","heyglif","Flovaai","CharaspowerAI","azed_ai","pika_labs","runwayml","Kling_ai","AIWarper"]: tw = raw_all.get(h, []) sel = [] for t in tw: dt = pd(t.get("createdAt","")) if not dt or dt < CUT or not is_orig(t): continue sel.append((dt, t)) if not sel: print(f"\n @{h}: none in window"); continue print(f"\n @{h} ({len(sel)} in-window originals):") for dt,t in sorted(sel, key=lambda z: z[0], reverse=True): print(f" {dt.strftime('%m-%d %H:%M')}|{t.get('likeCount') or 0}L|{t.get('viewCount') or 0}V|{mediatype(t)}|{(t.get('text') or '')[:150]}")