@m13v/s4l 1.6.203-rc.1 → 1.6.203-rc.2

package/mcp/dist/index.js

@@ -1301,7 +1301,7 @@ tool("engagement_mode", {
             .optional()
             .describe("The RAW voice-memo transcript from the onboarding dictation interview, VERBATIM (do NOT " +
             "paraphrase or summarize). Persisted to the persona_corpus.txt sidecar (never config.json), " +
-            "capped ~8000 chars. This is the grounding pool the drafter quotes real specifics from " +
+            "capped ~100000 chars. This is the grounding pool the drafter quotes real specifics from " +
             "(actual projects, numbers, opinions, phrasing), so keep it dense and first-hand."),
         voice: z
             .any()
@@ -2962,6 +2962,12 @@ function kickerEnv() {
         S4L_STATE_DIR: sapsStateDir(),
         TWITTER_TAIL_LINK_RATE: "0",
         TWITTER_PAGE_GEN_RATE: "0",
+        // Rolling virality bar percentile for the draft lane (2026-07-03): the
+        // cycle script now applies the bar to DRAFT_ONLY too, so below-bar picks
+        // never become review cards. p0.90 (not the script's 0.97 default) keeps
+        // card volume useful while cutting bottom-of-pool drafts; cold start
+        // (sample_count < 200) still shows brand-new installs every draft.
+        S4L_TWITTER_VIRALITY_PCTILE: "0.90",
     };
 }
 async function ensureQueueKickerInstalled() {

package/mcp/dist/setup.js

@@ -399,7 +399,7 @@ export function ensurePersonaProject(grounding) {
     if (grounding && typeof grounding.content_corpus === "string" && grounding.content_corpus.trim()) {
         try {
             const corpusPath = path.join(path.dirname(cfgPath), "persona_corpus.txt");
-            fs.writeFileSync(corpusPath, grounding.content_corpus.trim().slice(0, 8000) + "\n", "utf-8");
+            fs.writeFileSync(corpusPath, grounding.content_corpus.trim().slice(0, 100000) + "\n", "utf-8");
         }
         catch {
             // ignore: corpus is grounding fuel, not required for a working persona.

package/mcp/dist/telemetry.js

@@ -205,7 +205,7 @@ function collectStateSnapshot() {
     const config = readJsonCapped(cfgPath, 512_000);
     if (config !== undefined)
         state.config = config;
-    const corpus = readTextCapped(path.join(path.dirname(cfgPath), "persona_corpus.txt"), 16_000);
+    const corpus = readTextCapped(path.join(path.dirname(cfgPath), "persona_corpus.txt"), 100_000);
     if (corpus !== undefined)
         state.persona_corpus = corpus;
     const stateFiles = [

package/mcp/dist/version.json

@@ -1,4 +1,4 @@
 {
-  "version": "1.6.203-rc.1",
-  "installedAt": "2026-07-04T02:38:15.358Z"
+  "version": "1.6.203-rc.2",
+  "installedAt": "2026-07-04T04:11:58.027Z"
 }

package/mcp/manifest.json

@@ -2,7 +2,7 @@
   "dxt_version": "0.1",
   "name": "social-autoposter",
   "display_name": "S4L",
-  "version": "1.6.203-rc.1",
+  "version": "1.6.203-rc.2",
   "description": "Draft, review, approve, and autopilot X/Twitter posts.",
   "long_description": "## **⚠️ The disclaimer above is generic Claude boilerplate.** Anthropic shows the same warning on every plugin regardless of what it does; any plugin has the same level of access as any app you download from the internet.\n\nS4L is an open source product developed by Mediar.ai Incorporated, a VC-backed San Francisco-based startup.\n\nTo get started:\n\n1\\. Copy this prompt: **Set me up on S4L plugin end to end**\n\n2\\. Quit with CMD+Q, reopen Claude, paste into a new chat.\n\nWhat happens next:\n\n* About every 5 minutes S4L scans X for posts that match your topics and drafts replies in your voice.\n* Drafts show up as review cards, usually the first within a few minutes. Nothing is posted automatically; you approve each one.\n* Posting autopilot stays off until you explicitly turn it on.",
   "author": {

package/mcp/menubar/s4l_card.py

@@ -13,19 +13,23 @@ Decision shape: {"n": int, "approved": bool, "loved": bool, "text": str,
 "reject_note": str|None, "interactions": [{"type": str, "ts": str}],
 "dwell_ms": int}
 
-Approving comes in two strengths: the plain Approve button, and the 😄 button
-right next to it, which approves AND stamps loved=True: the user's "this one
-was a really good pick" signal, which the feedback digest treats as strong
-positive evidence (a plain approve is merely counter-evidence against avoid
-entries).
-
-Overall feedback (guidance not tied to any single thread) has two entry
-points that share one submit handler, registered by the menu bar via
-`set_feedback_handler` (it ships decision='feedback' review events): the
-card's Feedback button swaps the card body for an in-window composer (a
-separate floating panel was tried and opened where the user never saw it on
-a multi-monitor setup), and the menu bar's "Send feedback…" item opens the
-standalone `present_feedback(on_submit)` panel (no card window need exist).
+Approving is ONE button with click-to-escalate strength (two side-by-side
+approve buttons were tried first and read as clutter, 2026-07-03 feedback):
+the first click arms a plain approve, and each extra click within the short
+commit window bumps the reaction (👍 → 😄 → ❤️‍🔥). The decision commits when
+the clicks stop. Any level past the first stamps loved=True: the user's
+"this one was a really good pick" signal, which the feedback digest treats
+as strong positive evidence (a plain approve is merely counter-evidence
+against avoid entries); the exact level rides along as an approve_level_N
+interaction.
+
+Overall feedback (guidance not tied to any single thread) lives in the MENU
+BAR only (the card had its own Feedback button + in-window composer, but it
+crowded the action row and was orthogonal to the per-draft decision, so it
+moved out, 2026-07-03 feedback): the menu bar's feedback item opens the
+standalone `present_feedback(on_submit)` panel, whose submit handler the
+menu bar registers at boot via `set_feedback_handler` (it ships
+decision='feedback' review events).
 
 Rejecting is a two-step flow: the Reject button swaps the card body for a
 reason picker (three one-tap categories plus Other, with an optional free-text
@@ -118,6 +122,16 @@ REJECT_REASONS = (
 # Client-side cap on tracked interactions per card (server clips at 50 too).
 MAX_INTERACTIONS = 50
 
+# One-button approve escalation: button title per armed level (index = level).
+# Level 1 = plain approve; 2+ = loved=True on the decision, with the exact
+# level shipped as an approve_level_N interaction for the feedback digest.
+APPROVE_TITLES = ("Approve", "Approve 👍", "Approve 😄", "Approve ❤️‍🔥")
+APPROVE_MAX_LEVEL = len(APPROVE_TITLES) - 1
+# How long after the last approve click the decision commits. Long enough to
+# land a second/third click, short enough that a single approve still feels
+# instant-ish.
+APPROVE_COMMIT_DELAY = 0.6
+
 # Review-surface state mirrored to the state dir for out-of-process observers
 # (the menu bar watchdog, the dashboard, a debugging session). In the 2026-07-02
 # incident a card sat unseen for 3 hours and NOTHING on disk could distinguish
@@ -358,7 +372,13 @@ class _ReviewController(NSObject):
         self._interactions = []
         self._card_shown_at = None
         self._reason_field = None
-        self._feedback_tv = None
+        # Armed-but-uncommitted approve level for the CURRENT card (0 = none).
+        # Doubles as the staleness token for the delayed commit: every
+        # scheduled commit carries the level it was armed with, and only the
+        # one matching the live level fires (reject/advance/close just zero
+        # the level to void all in-flight commits — no cancellation API).
+        self._approve_level = 0
+        self._approve_btn = None
         # Attention anchors for the unattended-review watchdog: the stack counts
         # as "touched" on present, on any tracked interaction, and on any
         # decision. No touch past the watchdog threshold = the user is not
@@ -532,52 +552,25 @@ class _ReviewController(NSObject):
             self._interactions = []
             self._card_shown_at = time.time()
         self._reason_field = None
+        self._approve_level = 0  # voids any in-flight delayed commit
         content = NSView.alloc().initWithFrame_(NSMakeRect(0, 0, W, H))
 
-        # Buttons at the TOP, one line: Approve, Approve 😄 (approve + loved),
-        # Feedback (overall feedback, decides nothing), Reject at the right.
-        # Widths are hand-fit so all four bezeled buttons share the 348pt row
-        # without truncating their titles.
-        approve = NSButton.alloc().initWithFrame_(NSMakeRect(M, H - 42, 78, 30))
-        approve.setTitle_("Approve")
+        # Buttons at the TOP, one line: a single Approve at the left (click
+        # again within the commit window for a stronger reaction), Reject at
+        # the right. Overall feedback moved to the menu bar item.
+        approve = NSButton.alloc().initWithFrame_(NSMakeRect(M, H - 42, 130, 30))
+        approve.setTitle_(APPROVE_TITLES[0])
         approve.setBezelStyle_(NSBezelStyleRounded)
         approve.setTarget_(self)
         approve.setAction_("approve:")
-        content.addSubview_(approve)
-
-        # "Approve 😄" = approve with the "really good one" signal (loved=True
-        # on the decision). Same posting path as Approve; only the feedback
-        # rail sees the difference. Labeled as an approve variant, not a bare
-        # emoji: a lone 😄 read as decoration and users doubted the click
-        # registered (2026-07-03 feedback).
-        smile = NSButton.alloc().initWithFrame_(NSMakeRect(M + 82, H - 42, 100, 30))
-        smile.setTitle_("Approve 😄")
-        smile.setBezelStyle_(NSBezelStyleRounded)
-        smile.setTarget_(self)
-        smile.setAction_("approveLoved:")
-        try:
-            smile.setToolTip_("Approve and mark as a really good pick")
-        except Exception:
-            pass
-        content.addSubview_(smile)
-
-        # Feedback = overall feedback (about the pipeline, not this draft).
-        # Swaps the card body for the composer IN THIS WINDOW; a separate
-        # floating panel was tried first and opened somewhere the user never
-        # saw on a multi-monitor setup (2026-07-03: 7 clicks, zero sightings).
-        # A borderless bubble.left icon was tried next and read as decoration,
-        # not a button; a labeled bezel button matching its row mates won
-        # (2026-07-03 feedback).
-        fb = NSButton.alloc().initWithFrame_(NSMakeRect(M + 186, H - 42, 90, 30))
-        fb.setTitle_("Feedback")
-        fb.setBezelStyle_(NSBezelStyleRounded)
-        fb.setTarget_(self)
-        fb.setAction_("feedbackOpen:")
         try:
-            fb.setToolTip_("Send overall feedback (not about this draft)")
+            approve.setToolTip_(
+                "Approve; click again for a stronger reaction (😄 → ❤️‍🔥)"
+            )
         except Exception:
             pass
-        content.addSubview_(fb)
+        content.addSubview_(approve)
+        self._approve_btn = approve
 
         reject = NSButton.alloc().initWithFrame_(NSMakeRect(W - M - 66, H - 42, 66, 30))
         reject.setTitle_("Reject")
@@ -736,8 +729,11 @@ class _ReviewController(NSObject):
         self._textview = tv
 
         self._panel.setContentView_(content)
-        # Counter lives in the native title bar, not inside the content.
-        self._panel.setTitle_(f"Review draft {self._idx + 1} of {len(self._drafts)}")
+        # Counter lives in the native title bar, not inside the content, with
+        # the product name so a stray card is identifiable at a glance.
+        self._panel.setTitle_(
+            f"s4l · Review draft {self._idx + 1} of {len(self._drafts)}"
+        )
         # setContentView_ rebuilds the view tree, so the caret would otherwise
         # default to the Approve button. Re-seat it in the reply field for every
         # card (not just the first) so each one is immediately editable.
@@ -947,7 +943,7 @@ class _ReviewController(NSObject):
         # reset the caret / clobber an in-progress edit on the current card).
         try:
             self._panel.setTitle_(
-                f"Review draft {self._idx + 1} of {len(self._drafts)}"
+                f"s4l · Review draft {self._idx + 1} of {len(self._drafts)}"
             )
         except Exception:
             pass
@@ -970,112 +966,80 @@ class _ReviewController(NSObject):
 
     # ObjC selectors (trailing underscore -> "approve:" etc.)
     def approve_(self, sender):
-        self._record(True)
-        self._fire_decision()
-        self._advance()
-
-    def approveLoved_(self, sender):
-        _log("approved with love")
-        self._record(True, loved=True)
-        self._fire_decision()
-        self._advance()
-
-    def feedbackOpen_(self, sender):
-        # Overall feedback is orthogonal to the card decision: swap the card
-        # body for the composer (same pattern as the reject reason picker),
-        # preserving any in-progress reply edit for Back. In THIS window on
-        # purpose: a separate floating panel opened somewhere the user never
-        # saw on a multi-monitor setup.
-        self._track("feedback_open")
-        self._pending_reply_text = self._current_text()
-        self._render_feedback()
-
-    @objc.python_method
-    def _render_feedback(self):
-        content = NSView.alloc().initWithFrame_(NSMakeRect(0, 0, W, H))
-        content.addSubview_(
-            _label(NSMakeRect(M, H - 46, W - 2 * M, 20), "Overall feedback", size=13, bold=True)
-        )
-        content.addSubview_(
-            _label(
-                NSMakeRect(M, H - 80, W - 2 * M, 32),
-                "Standing guidance for the drafting loop (thread choice, tone, "
-                "audience), not about this draft.",
-                size=11,
-                muted=True,
-            )
+        # One button, click-to-escalate: every click bumps the armed level
+        # (capped) and restarts the commit window. Each scheduled commit
+        # carries the level it was armed with; only the one still matching
+        # self._approve_level when it fires actually commits, so extra
+        # clicks, reject, advance, and close all void in-flight commits by
+        # changing/zeroing the level.
+        if self._approve_level < APPROVE_MAX_LEVEL:
+            self._approve_level += 1
+        level = self._approve_level
+        try:
+            self._approve_btn.setTitle_(APPROVE_TITLES[level])
+        except Exception:
+            pass
+        _log(f"approve armed at level {level}")
+        self.performSelector_withObject_afterDelay_(
+            "commitApprove:", level, APPROVE_COMMIT_DELAY
         )
-        scroll, tv = _editable_scroll(NSMakeRect(M, 54, W - 2 * M, H - 86 - 54))
-        content.addSubview_(scroll)
-        self._feedback_tv = tv
-
-        back = NSButton.alloc().initWithFrame_(NSMakeRect(M, 14, 90, 30))
-        back.setTitle_("Back")
-        back.setBezelStyle_(NSBezelStyleRounded)
-        back.setTarget_(self)
-        back.setAction_("cardFeedbackBack:")
-        content.addSubview_(back)
-
-        send = NSButton.alloc().initWithFrame_(NSMakeRect(W - M - 90, 14, 90, 30))
-        send.setTitle_("Send")
-        send.setBezelStyle_(NSBezelStyleRounded)
-        send.setTarget_(self)
-        send.setAction_("cardFeedbackSend:")
-        content.addSubview_(send)
-
-        self._textview = None
-        self._panel.setContentView_(content)
-        self._panel.makeFirstResponder_(tv)
-        _log("in-card feedback composer shown")
 
     @objc.python_method
-    def _restore_card(self):
-        # Re-render the card and restore any reply edit made before the swap
-        # (same mechanism as rejectBack_).
-        pending = getattr(self, "_pending_reply_text", None)
-        self._feedback_tv = None
-        self._render()
-        if pending is not None and self._textview is not None:
-            try:
-                self._textview.setString_(pending)
-            except Exception:
-                pass
+    def _commit_approve(self, level):
+        """Record the armed approve. Level 2+ is the loved signal; the exact
+        strength rides along as an approve_level_N interaction."""
+        if level > 1:
+            self._track(f"approve_level_{level}")
+        self._record(True, loved=level > 1)
+        self._fire_decision()
 
-    def cardFeedbackSend_(self, sender):
-        text = ""
+    def commitApprove_(self, armed):
         try:
-            text = str(self._feedback_tv.string()).strip()
+            armed = int(armed)
         except Exception:
-            pass
-        if not text:
-            _log("in-card feedback sent empty; dropped")
-        else:
-            _log(f"in-card feedback submitted ({len(text)} chars)")
-            if _feedback_handler is not None:
-                try:
-                    _feedback_handler(text)
-                except Exception:
-                    pass
-        self._restore_card()
-
-    def cardFeedbackBack_(self, sender):
-        _log("in-card feedback cancelled")
-        self._restore_card()
+            return
+        if armed == 0 or armed != self._approve_level:
+            return  # stale: escalated further, rejected, advanced, or closed
+        self._approve_level = 0
+        self._commit_approve(armed)
+        self._advance()
 
     def reject_(self, sender):
         # Two-step reject: swap the card body for the reason picker. The
-        # decision is recorded when a reason chip (or Skip) is clicked. Any
-        # in-progress edit of the reply is preserved for Back.
+        # decision is recorded when a reason chip (or the no-reason button)
+        # is clicked. Any in-progress edit of the reply is preserved for
+        # Back. Voids an armed-but-uncommitted approve (last click wins).
+        self._approve_level = 0
         self._pending_reply_text = self._current_text()
         self._render_reason()
 
     @objc.python_method
     def _render_reason(self):
         content = NSView.alloc().initWithFrame_(NSMakeRect(0, 0, W, H))
+
+        # Buttons at the TOP, above the title, mirroring the card's action
+        # row so the cursor doesn't have to travel between steps: "Reject,
+        # no reason" sits right-aligned exactly where Reject was (one-click
+        # double-tap keeps the old zero-friction reject), Back sits at the
+        # left where Approve was.
+        back = NSButton.alloc().initWithFrame_(NSMakeRect(M, H - 42, 90, 30))
+        back.setTitle_("Back")
+        back.setBezelStyle_(NSBezelStyleRounded)
+        back.setTarget_(self)
+        back.setAction_("rejectBack:")
+        content.addSubview_(back)
+
+        skip = NSButton.alloc().initWithFrame_(NSMakeRect(W - M - 150, H - 42, 150, 30))
+        skip.setTitle_("Reject, no reason")
+        skip.setBezelStyle_(NSBezelStyleRounded)
+        skip.setTarget_(self)
+        skip.setAction_("rejectSkip:")
+        content.addSubview_(skip)
+
         content.addSubview_(
-            _label(NSMakeRect(M, H - 46, W - 2 * M, 20), "Why reject this draft?", size=13, bold=True)
+            _label(NSMakeRect(M, H - 72, W - 2 * M, 20), "Why reject this draft?", size=13, bold=True)
         )
-        y = H - 82
+        y = H - 106
         for i, (_, title) in enumerate(REJECT_REASONS):
             btn = NSButton.alloc().initWithFrame_(NSMakeRect(M, y, W - 2 * M, 28))
             btn.setTitle_(title)
@@ -1085,7 +1049,7 @@ class _ReviewController(NSObject):
             btn.setAction_("rejectReason:")
             content.addSubview_(btn)
             y -= 32
-        note = NSTextField.alloc().initWithFrame_(NSMakeRect(M, 56, W - 2 * M, 48))
+        note = NSTextField.alloc().initWithFrame_(NSMakeRect(M, 14, W - 2 * M, y + 28 - 14 - 8))
         note.setEditable_(True)
         note.setBezeled_(True)
         note.setFont_(NSFont.systemFontOfSize_(12))
@@ -1097,20 +1061,6 @@ class _ReviewController(NSObject):
         content.addSubview_(note)
         self._reason_field = note
 
-        back = NSButton.alloc().initWithFrame_(NSMakeRect(M, 14, 90, 30))
-        back.setTitle_("Back")
-        back.setBezelStyle_(NSBezelStyleRounded)
-        back.setTarget_(self)
-        back.setAction_("rejectBack:")
-        content.addSubview_(back)
-
-        skip = NSButton.alloc().initWithFrame_(NSMakeRect(W - M - 150, 14, 150, 30))
-        skip.setTitle_("Reject (no reason)")
-        skip.setBezelStyle_(NSBezelStyleRounded)
-        skip.setTarget_(self)
-        skip.setAction_("rejectSkip:")
-        content.addSubview_(skip)
-
         self._textview = None
         self._panel.setContentView_(content)
         self._panel.makeFirstResponder_(note)
@@ -1150,7 +1100,13 @@ class _ReviewController(NSObject):
 
     def windowShouldClose_(self, sender):
         # Closing the window stops review; remaining cards are left undecided
-        # (not posted). Finish with whatever was decided so far.
+        # (not posted). An approve armed within the last commit window is
+        # committed first — the user DID click Approve; losing it to a fast
+        # close would silently drop a decision. Finish with whatever was
+        # decided so far.
+        if self._approve_level:
+            level, self._approve_level = self._approve_level, 0
+            self._commit_approve(level)
         self._finish()
         return True
 
@@ -1263,11 +1219,11 @@ def heal_active():
 # ---- overall-feedback composer ----------------------------------------------
 # One small floating panel with a free-text field, for guidance that is about
 # the PIPELINE rather than any single draft ("less shilling", "more dev
-# threads", ...). Reachable from the card's Feedback button and the menu bar's
-# "Send feedback…" item; both call present_feedback(), which falls back to the
-# handler the menu bar registered at boot via set_feedback_handler() (that
-# handler ships a decision='feedback' review event down the same outbox rail
-# as card decisions, so the digest processes it the same way).
+# threads", ...). Reachable ONLY from the menu bar's feedback item, which
+# calls present_feedback(); that falls back to the handler the menu bar
+# registered at boot via set_feedback_handler() (the handler ships a
+# decision='feedback' review event down the same outbox rail as card
+# decisions, so the digest processes it the same way).
 
 FB_W = 380
 FB_H = 200

package/mcp/menubar/s4l_menubar.py

@@ -1999,8 +1999,9 @@ class S4LMenuBar(rumps.App):
         try:
             import s4l_card
 
-            # The card's 💬 (overall feedback) button opens the composer via the
-            # module-level default handler; register ours before any card shows.
+            # present_feedback (the menu bar's feedback item) falls back to
+            # the module-level default handler; register ours before any
+            # card shows.
             s4l_card.set_feedback_handler(self._on_feedback_text)
             s4l_card.present_review(
                 drafts,
@@ -2122,9 +2123,10 @@ class S4LMenuBar(rumps.App):
             pass
 
     def _on_feedback_text(self, text):
-        """Ship overall feedback (the card's 💬 button or the menu bar's
-        "Send feedback…" item) as a decision='feedback' review event on the
-        same outbox rail as card decisions. project is intentionally omitted
+        """Ship overall feedback (the menu bar's "Give feedback…" item; the
+        card had its own button once but it moved out, 2026-07-03 feedback)
+        as a decision='feedback' review event on the same outbox rail as
+        card decisions. project is intentionally omitted
         (NULL server-side): the feedback digest folds project-less feedback
         into EVERY configured project's prompt."""
         body = (text or "").strip()[:2000]
@@ -2386,7 +2388,11 @@ class S4LMenuBar(rumps.App):
 
         items.append(rumps.separator)
         items.append(rumps.MenuItem("Open dashboard", callback=self._open_dashboard))
-        items.append(rumps.MenuItem("Send feedback…", callback=self._menu_feedback))
+        # The one entry point for overall feedback (the review card no longer
+        # carries a Feedback button); named for what it does to the pipeline,
+        # not the mechanism.
+        items.append(rumps.MenuItem("Give feedback…", callback=self._menu_feedback))
+        items.append(self._label("   overall guidance, steers future drafts"))
         # While the update-verify marker is pending, the pipeline copy still
         # resolves the OLD version (it only advances once the restarted server
         # re-provisions repo/package, ~2 min), so the snapshot honestly reports

package/mcp/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@m13v/s4l-mcp",
-  "version": "1.6.203-rc.1",
+  "version": "1.6.203-rc.2",
   "private": true,
   "description": "Desktop MCP client for social-autoposter (X/Twitter rail): manual draft/review/approve loop, autopilot control, and stats. Thin wrapper over the existing pipeline scripts.",
   "license": "MIT",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@m13v/s4l",
-  "version": "1.6.203-rc.1",
+  "version": "1.6.203-rc.2",
   "description": "Automated social posting pipeline for Reddit, X/Twitter, LinkedIn, and Moltbook. Install as a Claude Code agent skill.",
   "bin": {
     "social-autoposter": "bin/cli.js",

package/scripts/build_persona.py

@@ -98,7 +98,7 @@ GROUNDING_INSTRUCTIONS = (
     "corpus as one plain-text block: the persona's actual posts, replies, repo "
     "descriptions, site copy, verbatim. This is NOT synthesized. It becomes the "
     "grounding pool the drafter quotes real specifics from, so keep it dense and "
-    "first-hand. Trim only obvious noise; do NOT paraphrase. Cap ~8000 chars.\n"
+    "first-hand. Trim only obvious noise; do NOT paraphrase. Cap ~100000 chars.\n"
     + PUBLIC_ONLY_NOTE
     + "\nThen write the fields to /tmp/persona.json and run "
     "`build_persona.py apply --from /tmp/persona.json`, or hand them to the "
@@ -346,7 +346,7 @@ def cmd_apply(args) -> int:
     if isinstance(corpus, str) and corpus.strip():
         corpus_path = cfg_path.parent / "persona_corpus.txt"
         try:
-            corpus_path.write_text(corpus.strip()[:8000] + "\n")
+            corpus_path.write_text(corpus.strip()[:100000] + "\n")
             corpus_written = str(corpus_path)
             changed.append("content_corpus")
         except Exception as e:

package/scripts/feedback_digest.py

@@ -18,12 +18,14 @@ scripts/learned_preferences.py for the full loop). Per run:
      never digested twice. Events are marked processed even when the plan is
      "no changes" — a considered no-op is a completed digestion, not a retry.
 
-Overall feedback (decision='feedback', project IS NULL; typed into the card's
-💬 composer or the menu bar's "Send feedback…" item) is fetched once per run,
+Overall feedback (decision='feedback', project IS NULL; typed into the
+composer behind the menu bar's "Give feedback…" item) is fetched once per run,
 folded into EVERY configured project's prompt as explicit standing guidance,
 and marked processed only after all attempted project digests succeed.
-Loved approvals (the card's 😄 button) arrive as loved=true on approved
-events and are surfaced to the model as strong positive evidence.
+Loved approvals (the card's escalated approve: extra clicks on the one
+Approve button) arrive as loved=true on approved events, with the exact
+strength as an approve_level_N interaction, and are surfaced to the model
+as strong positive evidence.
 
 Failure handling: a Claude failure or unparseable plan leaves the events
 unprocessed for the next run. A run-level flock prevents concurrent digests.
@@ -163,7 +165,7 @@ CURRENT content_guardrails.do_not: {json.dumps(guard_do_not)}
 NEW REVIEW EVENTS since the last digest ({len(rejected)} rejected, {len(approved)} approved, {len(loved)} of the approvals loved):
 {ev_lines}{overall_block}
 
-Categories: wrong_author = the thread's author/audience was a bad fit; off_topic = the thread itself was a bad fit; bad_draft = thread was fine but the written reply was off; other = see the note. "user_checked=profile_click" means the user opened the author's profile before deciding (a strong author-quality signal even without a note). "[approved+loved]" means the user pressed the emphatic-approve button ("this was a really good one"): strong positive evidence for audience_prefer and thread selection, worth roughly two plain approvals.
+Categories: wrong_author = the thread's author/audience was a bad fit; off_topic = the thread itself was a bad fit; bad_draft = thread was fine but the written reply was off; other = see the note. "user_checked=profile_click" means the user opened the author's profile before deciding (a strong author-quality signal even without a note). "[approved+loved]" means the user clicked Approve extra times for a stronger reaction ("this was a really good one"; approve_level_N in interactions carries the strength): strong positive evidence for audience_prefer and thread selection, worth roughly two plain approvals.
 
 You can also block SPECIFIC authors via the plan's block_authors list. A block is a permanent hard exclusion of that one handle from all future thread selection, so it is YOUR judgment call, never automatic. Block when the evidence is strong: a wrong_author reject IS a direct human statement about that author (especially with profile_click), and the author context (author_followers, their post, found_via_topic) or the user's note confirms the account itself was the problem rather than the topic. Do NOT block when the reject looks topic-driven (off_topic/bad_draft on a reasonable account) or when you are unsure; the generalizable TYPE entry in audience_avoid is the softer tool for that.
 

package/scripts/pick_search_topic.py

@@ -570,24 +570,29 @@ def pick_topic_for_project(project_name, platform="twitter",
         for i in range(min(REFERENCE_TOP_N, len(pool)))
     ]
 
-    # 2026-05-28 dedicated explore branch for freshly-invented topics.
-    # 10% of the time, if there's any source='invented' + attempts_n==0
-    # topic in the pool, pick one of those uniformly. Gives invent_topics.py
-    # outputs a guaranteed sampling cadence (~1 sample every other day per
-    # project at default 5 picks/day) instead of languishing at the
-    # COLD_TOPIC_WEIGHT floor for weeks. Falls through to the normal
-    # weighted-random branch when (a) no eligible invented-untried topics
-    # exist OR (b) the random roll lands outside the explore rate.
+    # 2026-05-28 dedicated explore branch for never-tried topics.
+    # 10% of the time, if there's any attempts_n==0 topic in the pool
+    # (regardless of source: invented by invent_topics.py OR freshly
+    # seeded/appended into config.json by a human or agent), pick one of
+    # those uniformly. Gives every brand-new topic a guaranteed sampling
+    # cadence (~1 sample every other day per project at default 5 picks/day)
+    # instead of languishing at the COLD_TOPIC_WEIGHT floor for weeks.
+    # 2026-07-04: widened from source=='invented' to any attempts_n==0 so a
+    # topic added to config after bootstrap (behaviorally identical to an
+    # invention: new, untried, needs its first shot) gets the same first-shot
+    # lane and proves itself on real click/supply signal from there. Falls
+    # through to the normal weighted-random branch when (a) no eligible
+    # untried topics exist OR (b) the random roll lands outside the rate.
     invented_untried_idxs = [
         i for i, r in enumerate(pool)
-        if r.get("source") == "invented"
-        and int(r.get("attempts_n") or 0) == 0
+        if int(r.get("attempts_n") or 0) == 0
     ]
     if invented_untried_idxs and rnd.random() < INVENTED_UNTRIED_EXPLORE_RATE:
         chosen_idx = rnd.choice(invented_untried_idxs)
         sys.stderr.write(
-            f"[pick_search_topic] invented_untried_explore "
+            f"[pick_search_topic] untried_explore "
             f"project={project_name!r} pool_size={len(invented_untried_idxs)} "
+            f"source={pool[chosen_idx].get('source')!r} "
             f"chosen={pool[chosen_idx]['search_topic']!r}\n"
         )
         chosen = pool[chosen_idx]

package/skill/run-twitter-cycle.sh

@@ -2128,19 +2128,23 @@ POST_TOP_N="${S4L_TWITTER_POST_TOP_N:-1}"
 # (~20-30 / 8h) with NO hard cap: the bar is the Nth percentile of the install's
 # OWN recent candidate pool (via /api/v1/twitter-candidates/virality-threshold),
 # so it self-calibrates to cadence and niche instead of being a fixed number.
+# The bar applies to BOTH lanes (2026-07-03, per user instruction): the
+# autopilot lane drops below-bar picks before POSTING, and the DRAFT_ONLY lane
+# drops them before they become review cards, so human review time is never
+# spent on bottom-of-pool drafts. Dropped picks stay status='pending' (never
+# 'rejected'); Phase 0 salvages and re-judges them next cycle.
 # The bar is OFF (empty threshold) when:
-#   - DRAFT_ONLY: new users / manual review see every draft (we don't even fetch).
 #   - Cold start: sample_count < min, so a fresh pool posts ungated until it fills.
+#     (This is also what keeps brand-new installs seeing every draft card.)
 #   - Fetch failure: fail-open, never silence posting on a transient API blip.
 # Tunables: S4L_TWITTER_VIRALITY_PCTILE (default 0.97),
 #           S4L_TWITTER_VIRALITY_MIN_SAMPLE (default 200).
-VIRALITY_THRESHOLD=""
-if [ "${DRAFT_ONLY:-0}" != "1" ]; then
-    VIRALITY_THRESHOLD=$(S4L_VPCTILE="${S4L_TWITTER_VIRALITY_PCTILE:-0.97}" \
-        S4L_VMIN="${S4L_TWITTER_VIRALITY_MIN_SAMPLE:-200}" \
-        python3 -c "
+VIRALITY_THRESHOLD=$(S4L_VPCTILE="${S4L_TWITTER_VIRALITY_PCTILE:-0.97}" \
+    S4L_VMIN="${S4L_TWITTER_VIRALITY_MIN_SAMPLE:-200}" \
+    S4L_SCRIPTS_DIR="$REPO_DIR/scripts" \
+    python3 -c "
 import os, sys
-sys.path.insert(0, os.path.expanduser('~/social-autoposter/scripts'))
+sys.path.insert(0, os.environ.get('S4L_SCRIPTS_DIR') or os.path.expanduser('~/social-autoposter/scripts'))
 from http_api import api_get
 try:
     r = api_get('/api/v1/twitter-candidates/virality-threshold',
@@ -2154,11 +2158,10 @@ try:
 except BaseException as e:
     sys.stderr.write(f'virality-bar fetch failed (bar OFF this cycle): {e}\n')
 " 2>>"$LOG_FILE" || echo "")
-    if [ -n "$VIRALITY_THRESHOLD" ]; then
-        log "Virality bar ACTIVE: p${S4L_TWITTER_VIRALITY_PCTILE:-0.97} = $VIRALITY_THRESHOLD (this install, trailing 24h); top-1 posts only if it clears the bar."
-    else
-        log "Virality bar OFF this cycle (cold-start/thin pool or fetch failed); top-1 posts ungated."
-    fi
+if [ -n "$VIRALITY_THRESHOLD" ]; then
+    log "Virality bar ACTIVE: p${S4L_TWITTER_VIRALITY_PCTILE:-0.97} = $VIRALITY_THRESHOLD (this install, trailing 24h); top-1 kept only if it clears the bar."
+else
+    log "Virality bar OFF this cycle (cold-start/thin pool or fetch failed); top-1 kept ungated."
 fi
 
 # Parse the prep envelope and write the plan to \$PLAN_FILE; also extract the