Then I hit a wall with /ticket-pickup. It was stopping after the plan phase instead of spawning the worktree lane. Felt stupid the second I saw it — the command had all the right pieces, but the execution flow was backwards. Fixed it to actually create the lane after planning. That's the kind of thing that burns time because everything looks right until you trace through what actually runs.

The bigger refactor was consolidating the agent-board registry system. Had three different places storing session state, two of them barely used. Merged them into a single session registry that the cockpit can query, and pushed worktree-only lanes to use that instead of the old ad-hoc naming. This cut about 40 lines of conditional logic from agent-board.sh, but now I'm worried about the edge case where a session gets registered but the worktree never actually creates. Haven't hit it yet, but the cleanup path isn't obvious if it does.

Also burned time fixing shell script permissions and consolidating install logic across install.sh, install-mac.sh, and the symlink rewiring. The dotfiles repo has grown into a mess of overlapping responsibilities — hooks, tmux configs, zsh setup, Claude agent definitions all living in the same tree. Works fine until it doesn't, and then you're debugging why a symlink didn't resolve or a script didn't execute.

Split the Slack TLDR view into tabs now. Alert channels ring the bell, monitor channels are silent. Added a test channel to the config to keep alerts from going dark when nobody's watching. Arrow keys tab between views, which is cleaner than the old modal switching.

Still need to wire up the /scope command to actually use conversation context instead of just dumping raw ticket details. And the ticket-pickup type detection is basic — it just looks at keywords in the title. That'll probably break on something weird.

This week: Paris closes out a Space Invaders sub-series with two fresh additions (PA_1569 and PA_1570), while London, New York, and Fontainebleau take hits with three destructions logged in as many days. A quieter week on the press front, but the city-by-city pulse tells its own story.

Paris caps a series, and the rest of the map takes a beating

The headline news this week comes out of Paris, where Invader added PA_1569 and PA_1570 — explicitly flagged in the event feed as "les deux derniers SI parisiens," the last two pieces in a Parisian Space Invaders sub-series. For a city that already dominates the global map with 48 events tracked (nearly triple any other city), capping a series like this feels less like routine maintenance and more like a punctuation mark. Hunters in the 11th — and wherever PA_1569 and PA_1570 actually landed — now have two fresh targets to chase before someone else gets the first flash. Paris also saw a status update on PA_1275, the kind of housekeeping note that usually signals damage, repair, or a change in visibility worth checking in person.

The same week, however, was brutal elsewhere. London lost LDN_129 to degradation on the 7th, and the 8th brought a double blow with destructions of FTBL_36 in Fontainebleau and NY_176 in New York. Three cities, three pieces gone or compromised inside 48 hours. It's a sharp reminder of what makes this whole project so emotionally charged: every mosaic out there is borrowed time. The ceramic-tile-as-permanent-medium idea is a beautiful one, but the street always gets the final say. If you've got an un-flashed invader sitting on a "next trip" list, this is your nudge.

The shape of the map right now

Zooming out from the week's events, the broader activity map continues to reinforce what regulars already feel intuitively: Paris is the heart, Europe is the body, and the rest of the world is a constellation of passionate outposts. Paris alone accounts for roughly 35% of documented global events, with Bern and London tied at 11 events apiece as secondary European hubs. Across the Atlantic, Los Angeles (13) and New York (9) anchor North America, while São Paulo (12) holds down South America almost single-handedly and Hong Kong (3) flies the flag for an Asia-Pacific scene that's been quieter than its history suggests. This week's destructions hitting London, NYC, and Fontainebleau all land squarely in those high-activity zones — which is exactly where you'd expect attrition to bite hardest, since more pieces and more foot traffic mean more wear, more cleanup crews, and more opportunistic removals.

Worth a click

Curated sources were largely static this week, but the open-web surfaced a piece worth bookmarking for anyone deep in the lore: GraffitiStreet's coverage of Invader's nineteenth London invasion wave, here, is a useful companion read given that London just lost LDN_129 — context for how the city's stock has been built up wave by wave, even as individual pieces fall. MyArtBroker's perpetual guide to the top 10 places to find an Invader original also resurfaced in this week's deltas, which, taken together with the Paris series-cap, makes for a fitting week to think about the geography of this whole project: where it concentrates, where it thins out, and where it's quietly disappearing.

This week: Paris closes out its latest invasion wave with two fresh mosaics (PA_1569 and PA_1570), while hunters mourn losses in Fontainebleau, New York, and London. The street keeps giving, and the street keeps taking.

Paris Caps Off a Wave

The big news this week comes straight from the capital: the spotter feed confirms the addition of PA_1569 and PA_1570, described as the last two Space Invaders of the current Parisian series. There's a real sense of occasion to a moment like this — the closing punctuation on a wave that hunters have been chasing piece by piece. Paris already dominates the global map with roughly three times the activity of any other city, and these two new tiles push that lead even further. If you're planning a Paris trip in the coming weeks, your checklist just got two entries longer. Meanwhile, the same feed logged a status update on PA_1275, a small reminder that even the older pieces are living, breathing things in the database — conditions change, and the catalog moves with them.

Three Cities Lose a Piece

It wasn't a clean week. Fontainebleau lost FTBL_36, New York lost NY_176, and London saw LDN_129 degraded — three separate strikes across three countries in a matter of days. None of these are surprising in the abstract; ceramic on a street wall is always negotiating with weather, construction crews, taggers, and the occasional collector with a chisel. But each loss lands a little harder when you've flashed the piece yourself, or planned to. NY_176 in particular feels notable given New York's relatively modest count of nine documented events on the global map — every Manhattan or Brooklyn loss thins the herd more than a Parisian one does. For anyone with a trip to London or Fontainebleau on the calendar, it's worth checking your local maps before you walk a route built around a piece that's no longer fully there.

The Wider Conversation

On the press and fan-source side, the curated feeds at MyArtBroker (both the artist page and the evergreen Top 10 Places to Find an Invader Original guide) refreshed this week, alongside the official Invader Instagram and the ever-essential invader-spotter news page. The MyArtBroker guide remains a solid primer for anyone new to the hunt — it leans heavily on Rue Monge in the 5th Arrondissement as a pilgrimage spot, which feels especially fitting in a week when Paris is the story. Open-web crawls also resurfaced GraffitiStreet's coverage of the nineteenth London wave and Artnet's piece marking Invader passing 4,000 mosaics — useful context as the project keeps barreling past its own milestones.

The Rhythm of the Game

Two added, three lost or degraded, one status update — a net-negative week on paper, but that's the deal we signed up for. The pieces aren't supposed to last forever; they're supposed to be found, photographed, argued about, and eventually replaced by something new on a wall somewhere across town. PA_1569 and PA_1570 are out there right now waiting for their first flashes. Go look up.

Then pivoted to the status display — two separate changes to keep the agent board readable. Dropped the AGE and PORT columns from the tmux board because they were noise; contexts that don't have any work output now get suppressed entirely instead of printing blank lines. Keeps the eye focused on what's actually running.

The pacing meter got more interesting. Was showing a simple daily quota percentage, but that doesn't say much about momentum. Rewrote it to track a 7-day delta — how many stints completed in the rolling window versus the target. Then converted that to a ratio that shows whether you're on pace, ahead, or behind. It's a single number (0.87) that means more than a percentage. The shell script's ugly now because I'm computing this in awk instead of delegating to the actual service, but it was faster to prototype here.

What's unclear: whether the HTTP endpoint should be read-only or if we eventually need to trigger digest regeneration from the client. Right now it's just fetching whatever the background worker cached. Also haven't decided if the pacing ratio should factor in the type of work or treat all stints equally. Some days you need deep focus, some days you're in meetings. The math is agnostic but the meaning shifts.

This week: Paris closed the chapter on its current invasion wave with the arrival of PA_1569 and PA_1570, while four mosaics fell to destruction across New York, London, Melbourne, and Paris itself. London also saw LDN_178 quietly reactivated, a small win against a week of attrition.

Two Final Parisians, and a Week of Losses

The headline this week is unmistakably Parisian: on the 7th, PA_1569 and PA_1570 were added — flagged in the spotter feed as "les deux derniers SI parisiens," the closing pair of the current Paris sequence. For a city that already accounts for roughly a third of global activity on the map, watching the numbering tick up to 1,570 is the kind of milestone that puts the scale of this project into sharp relief. Paris has been the gravitational center of Invader's work since the 1990s, and rounding out another batch is always a moment worth pausing on, especially for hunters who'll be plotting routes through the arrondissements this spring.

The other side of the ledger was harsher. Four destructions in a single week — PA_1223 in Paris on the 3rd, MLB_19 in Melbourne on the 4th, LDN_129 in London on the 7th (logged as a dégradation rather than a clean destruction, but battered all the same), and NY_176 in New York on the 8th — span four cities and three continents. That's the unsentimental rhythm of street art: ceramic is durable, but cities are not gentle. Anyone within range of NY_176 or PA_1223 who hasn't flashed them yet can consider those particular squares of the gallery permanently closed. One quiet bright spot: LDN_178 was reactivated on the 4th after a status update, a reminder that the spotter community's vigilance occasionally rescues a piece from the "lost" column.

A City-by-City Pulse

The week's events map cleanly onto the broader global picture. Paris leads all activity on the world map with 48 logged events, and this week alone it accounted for two of the three additions and one destruction — fully half the action. London (11 events globally) contributed both a degradation and a reactivation, a microcosm of how mature invasion zones tend to behave: lots of churn, with pieces flickering between active, damaged, and restored states. New York (9 events) and Melbourne round out the geographic spread. For hunters chasing city completion, weeks like this are a useful nudge to prioritize the older, more vulnerable IDs before weather, construction, or collectors get to them first.

From the Wider Web

On the curated-press side, MyArtBroker continues to maintain its Invader artist hub and its guide to the top ten places to find an Invader original (https://www.myartbroker.com/artist-invader and https://www.myartbroker.com/artist-invader/guides/top-10-places-to-find-an-invader-original) — both worth a bookmark if you're newer to the secondary market or planning a hunting trip around known invasion zones. The invader-spotter.art news feed (https://www.invader-spotter.art/news.php) remains, as ever, the single most reliable pulse-check for additions, destructions, and status changes; this week's events were all sourced there, and it's the page most of us refresh more often than we'd care to admit.

A final thought for the week: with PA_1569 and PA_1570 tucked into place and four pieces lost, the net is barely negative — but every destruction is a reminder that flashing is, in its small way, an act of preservation. Get out, look up, and stay safe out there.

The slack-tldr daemon is working, sort of. Got it blinking new alerts at 2 Hz using inverse-video frame toggling in slack-tldr.py, and the ack mechanism in slack-watch is solid. But the backfill logic is brittle. When you join a channel with ten thousand messages, the over-fetch (I'm pulling 50% more than asked) was meant to protect against join-spam starving the TLDR digest. It works, but it's a band-aid. I'm fetching too much, too often, and watching the config balloon as a workaround tells me the architecture wants something else.

The cockpit system — the worktree lanes + agent-state board + command suite — is real now. Prefix+L gives you a 4x2 tmux grid. /ticket-pickup pulls a ticket into your current lane state. But there's friction between how the state hooks (_state-write.sh, agent-state-active.sh) talk to the command layer. The refinement pass killed a lot of redundancy in the agent profiles (backend.md, frontend.md, etc.), but I kept adding /ticket-pickup and now I'm not sure if that's a command or a stateful operation or both.

Next is making the worktree protocol docs actually binding — right now it's guidance in worktree-protocol.md, but nobody's enforcing it. Also need to kill the token bleed on session start properly, not just trim symptoms.

The fix was dumb in hindsight: I added a small buffer queue in audio-watcher.py that holds pending renders while waiting for commit metadata. Nothing fancy. Just a deque with a 200ms window. It works, but now I'm carrying around this magic number that I'll probably regret in three months when the timing constraints change.

The actual feature I was chasing is audio-reactive rendering that pulls from multiple sources. Had to build out a config loader in audio-watcher.config.example.json that lets you specify different audio inputs—microphone, system output, whatever. Each source gets its own frequency analyzer thread. The commit-watcher.py module now emits branch events in addition to the usual commit signals, so the renderer can react differently when you switch branches mid-session. That part felt clean.

Wired it all together in watch.py and threw some shell aliases into .zshrc so I'm not typing python paths every time I want to spin up a session. The whole thing is... serviceable. Not elegant. The config format is JSON when it probably should've been YAML, and I'm doing string parsing in three different places that should be unified into a schema validator. But it works for what I'm building right now.

What I haven't solved: how to handle audio sources that drop frames or go silent for a few seconds. Right now the renderer just holds the last known state, which looks wrong. I could implement smoothing, interpolation, something. Instead I'm going to ship it broken and see if it actually matters when I'm using it in anger.

This week: Three mosaics fell to destruction across Melbourne, Paris, and a quieter zone, while London delivered both a degradation and a reactivation in the same 24-hour stretch — a reminder that the street giveth and the street taketh away.

A Rough Week on the Walls

The destruction column made for grim reading this week. MLB_19 in Melbourne came down on the 4th, PA_1223 was lost in Paris on the 3rd, and WN_04 met its end on the 29th in tandem with a degradation hit on LDN_21 across London. That London pairing is the kind of sequence that stings — one piece weathered, another erased, in the same breath. Paris losing PA_1223 is particularly notable given that the capital remains Invader's spiritual home and accounts for roughly a third of all global activity in our running map data; every Parisian loss chips at the densest gallery on Earth. Melbourne's MLB_19 also echoed strangely into a London status update later in the week, a small reminder of how interconnected the global ledger has become as hunters and trackers cross-reference cities in real time.

Reactivations Offer a Small Counterweight

Not all the news pulled in one direction. LDN_178 in London received a status update on the 4th, flipping back onto the active board, and BAB_42 was likewise reactivated on the 30th. Reactivations are one of the quieter joys of following Invader's catalogue — a piece thought lost or compromised gets reassessed, repaired, or simply re-confirmed in the wild, and suddenly the map looks a little fuller again. For BAB hunters in particular, BAB_42 returning to play is a nice mid-week lift in a city that sits firmly in the secondary activity tier with seven tracked events overall.

The Press Beat: MyArtBroker Spotlights the Hunt

On the curated press side, MyArtBroker refreshed two relevant pages this week: their main Invader artist hub at https://www.myartbroker.com/artist-invader and, more interestingly for hunters, their guide Top 10 Places to Find an Invader Original at https://www.myartbroker.com/artist-invader/guides/top-10-places-to-find-an-invader-original. The guide leans into Paris as the natural starting point, calling out Rue Monge in the 5th Arrondissement among its featured locations — a fitting nod given that Invader was born in Paris and has run his most concentrated invasion waves there. For newer flashers building a travel itinerary, pieces like this one are useful primers, even if seasoned hunters will already have walked Rue Monge a dozen times over.

Where the Map Stands

Stepping back from the week's individual events, the broader picture remains familiar: Paris leads the world in tracked activity by a wide margin, with London, Los Angeles, São Paulo, and Bern rounding out the upper tier. London's double-event week — a degradation on LDN_21 and a reactivation on LDN_178 — keeps it firmly in the conversation as one of the most dynamic non-Paris cities to hunt right now. And with BAB notching a reactivation of its own, the secondary tier continues to prove that the interesting stories aren't always in the headline cities. Look up, hunt safely, and as always — don't risk yourself for a flash.

Ripped out matrix.py entirely and wrote watch.py from scratch. The new renderer uses glyphs that fade in and out on a cycle, timed to meditation rather than machine urgency. Started at 12 seconds per breath, which felt frenetic. Jumped to 60 seconds. Still wrong. Then 180 seconds, and yeah—that's the cadence. The glyph gradient brightens at the inhale, dims at the exhale. It's subtle enough that you don't consciously watch it, but your brain syncs to it anyway.

Hit a wall with curses.COLORS. The function returns -1 if you call it before start_color, which I did. Spent forty minutes chasing a logic error that didn't exist—the bug was in the call order, not the math. That's the kind of thing that makes you feel stupid until you realize curses is just ancient and has opinions about initialization order that aren't documented anywhere obvious.

Also made commit-watcher.py respect the current working directory instead of hardcoding a repo path. Sounds trivial, but it means I can use the same script in multiple repos without symlink gymnastics. The $PWD lookup lives in the zsh hook now, passed as an argument on startup.

Separately, added a verification layer for Claude—a pre-flight check that sources any code claims against actual files before acting on them. Lives in verify-before-act.sh. The settings file documents the policy. I don't trust external systems making changes to my environment without at least asking what they're operating on.

The watch renderer isn't done. The glyph selection is still random, and I want each commit to have its own glyph that persists for a few cycles. Right now they're just noise that happens to be slow. Still working on making the visual language match the repository's actual heartbeat.

The dotfiles work is cleaner. Built a singleton file-lock pattern in commit-watcher.py so it doesn't spawn multiple instances when zsh boots. The watcher now auto-starts when you run art matrix — that command spawns the Matrix digital rain visualization, and the watcher sits in the background, feeding recent commits into the bottom rows of the display. Had to backfill the last N commits on boot or you'd get a blank state for the first few seconds, which felt broken. That's in commit-watcher.py around line 60-ish, reading from git log with a hardcoded window of 50 commits.

The life-os dashboard got gutted. Pulled out activities, alerts, agents, director — all the auxiliary views. Leaving news, flashcastr, and the blog. It's aggressive. People might land on the dashboard and wonder where half the features went, but the bloat was real. We're trying to build a command center, not a spreadsheet factory. The monitoring rule I added watches display_sync — if that table ever hits zero rows, we get an alert. That's the kind of thing that cascades quietly if you're not watching.

What's eating at me: the Docker startup chain still feels slow. I'm not sure if it's npm 11.x being heavier, or if the tsx invocation is doing extra work, or if it's the node process initialization itself. Haven't profiled it yet. The commit-watcher backfill is also arbitrary — why 50? Could be too much, could be too little depending on how fast you're committing. Haven't tested that assumption.

This week: The invasion sees mixed fortunes across Europe and beyond. While new life breathes into several dormant pieces through reactivations, the movement also suffered notable losses to degradation and destruction in London and Fontainebleau.

Reactivations Signal Active Maintenance

Four invaders returned to active status this week across three cities, suggesting Invader's ongoing commitment to refreshing his global catalog. In Grenoble, a cluster of three pieces—GNV_01, GNV_04, and GNV_11—received status updates on April 28, indicating either restoration work or renewed visibility after years of decline. Barcelona saw BAB_42 refreshed on April 30, while Brazil's BRL_09 was reactivated on April 25. These updates remind us that the invasion isn't static—it's a living, breathing project requiring constant attention and care.

Urban Losses: Destruction and Degradation

The week brought heartbreaking news from two fronts. On April 29, London's LDN_21 suffered degradation even as nearby WN_04 was destroyed entirely, a stark reminder that not all public spaces protect street art with equal commitment. Days earlier on April 26, Fontainebleau experienced a similar blow when FTBL_06 was destroyed following degradation of a nearby piece. These losses underscore the precarious existence of public art—constantly vulnerable to weather, urban renewal, and institutional indifference. Yet they also underscore why documenting and celebrating invaders through apps like FlashInvaders matters more than ever.

The notes-pipeline needed rewiring. Instead of feeding raw commit data to the generator and hoping it wouldn't confabulate, we now extract structured facts first—repo names, actual file lists, message content verbatim. A ticket-translation module maps Linear refs if they exist, but crucially, doesn't create them if they don't. The pipeline still generates drafts in pending-review, but now with a word-diff layer that flags where the output diverges from source material. That's imperfect and I hate it, but it catches the worst cases.

Wired up cross-agent enrichment via RPC. The agent-blog can now ask service-director and service-linear for context without hallucinating their responses. Built bus-rpc in shared/core to handle the back-and-forth. It's synchronous, which feels wrong for distributed systems, but we're not distributed yet—we're monorepo'd and the latency is negligible.

Got bitten by a partial unique index issue in the autonomous-tick handler. The ON CONFLICT clause wasn't matching because the index covered fewer columns than the insert. Spent an hour debugging why the same record kept getting reinserted. Fixed it in 0009_blog_stats_snapshots_uniq.sql but now I'm paranoid about other indexes doing the same thing silently.

Also spent time on terminal colors. Didn't plan to, but .tmux.conf and the new Alacritty config needed Gruvbox Material palettes to match the iTerm2 presets. Small win—environment consistency, but completely orthogonal to the core work.

Still unclear: whether the engagement learning loop in stats-ingestion is actually learning anything useful, or just recording noise. The preference-scores table exists, but I haven't built the feedback path yet. That's next.

The reconcile protocol got hairy. I kept adding weight parameters to emit() calls across five different agents—blog, email, farcaster, pet, rs3—because bulk imports were drowning out the progress signal. One agent would reconcile 200 transactions while another bumped a single calendar event, and the display-sync would think everything was done. Weighted reconciliation meant revisiting the shared base component and touching tsconfig across multiple workspaces.

Then there's the schema collapse. Display-sync and agent reconciliation used to speak different dialects. I collapsed them into a single protocol, which meant rewriting handlers across eight agents and a bunch of hotfixes when the drizzle journal timestamps went backwards (0029 wasn't past 0023). Fixed it with a journal bump, then fixed the actual bug where DateTime! scalar wasn't registered in admin typedefs.

Finance specifically ate two days. Built a 12-month narrative report that's cost-capped and goal-aware, added bucket categorization, rebuilt the entire ledger as raw CSV debit/credit instead of transaction-oriented storage. Parser now handles YYYYMMDD format. Wrote csv-parser.test fixtures to lock in the behavior. The annual report generates a forecast, but the forecast blob wasn't null-guarding statusQuo and the PWA would crash on stale data.

What's rough: the finance report generation is still fragile. Adding one more parser schema or changing how categories group takes rewriting the entire analysis pipeline. And I'm not sure the weighted reconcile is right—some agents emit fast and light, others slow and heavy. The scale's arbitrary right now.

Next is figuring out whether the forecast payload shape actually matches what the PWA expects.

Spent time on the image sync refactor, extracting Unsplash into service-image as its own component. Built out the routing-key changes and wired up agent-activity and agent-recipe to call into the new service instead of doing image fetches inline. The Railway deployment config is there (railway.toml), but I haven't tested the actual deployment flow yet—just set it up so the infra team can push it.

Calendar event dedup is half-working. Added a content_hash column to catch duplicate syncs within the 5-minute poll window, but if an event changes slightly (like description gets edited), we'll see it as new. I'm not sure if that's the right trade-off. The fallback to 5m polling when webhooks fail is solid, though.

Finance work got real messy. Amex exports charges as positive but our system expects negatives, so I had to negate them in the CSV parser. Deep-dive metrics were registering twice because the wrapper wasn't unwrapping the response correctly. Goals parser was choking on markdown formatting, so I hardened that too. The whole flow is still fragile—agent-finance expects the month and filename from the statement-uploaded event payload, but if that's malformed, the analysis silently fails.

Dotfiles got a major cleanup. Set up daily plan pruning via launchd, added a focus-guard tool (HTTP + HTTPS status page for blocked sites), and built out context-economy guardrails to track token usage. The statusline now shows the 5h/7d rate limit buckets when we're above 50%, which should give me better visibility into cost creep.

What's uncertain: whether the safety nets in display-sync are masking a deeper schema ordering problem, or if the calendar dedup strategy will survive real-world edits.

The bigger push was promoting agent-weather and agent-git from side agents to proper services with CONTEXT_REQUEST interfaces. Renamed the database column to process across both—three migration files per service to keep the old state intact during deploy. The schema safety nets I added in those migrations are doing real work now; without them, SQLite provisioning on startup fails silently and you get null columns everywhere.

Wired RS3 inferred activities into the dashboard widget, fixed the email KPI stats to actually show unread/synced counts, and expanded the news roster to 37 sources using a source adapter pattern in agent-news. The news page was crashing because articleIds stayed null—turned out the agent wasn't firing CONTENT_READY payloads with the right event shape.

Mobile layout work across recipes, gym, workout, jobs, football. The gutter token migration is cleaner now; grep-verified zero padding violations on page-container. Removed the grey box from the mobile upcoming fixture card.

Started seeing schema provisioning problems cascade hard. Added a startup probe in shared/rabbitmq/subscriber that hard-fails if the DB schema is missing expected columns—better than silent null returns. Also went through agent-activity, agent-calendar, agent-gym with schemaSafetyNets using PRAGMA table_info instead of IF NOT EXISTS, since older SQLite doesn't like the latter.

The notification service now hard-blocks web-push during sleep hours, pulling timezone config from shared. Per-widget ErrorBoundary on dashboard means one broken fetch doesn't crater the whole page.

What's unresolved: weather data is still stale sometimes—0,0 coordinates getting logged and the 15-min refresh isn't catching it fast enough. Also the activities page infinite reload loop on empty profile needs more investigation. Next is the recipe image URL rendering and activity reschedule mutations.

This week: The invasion experienced turbulence across multiple continents—destructions in Paris and New York, a crucial reactivation in an unnamed territory, and the complex interplay of creation and erasure that defines street art's precarious existence.

The Toll: Destructions and Degradations

The past seven days saw a sobering reminder that invasion survival remains uncertain. PA_300 in Paris fell to degradation on April 20, joining PA_213 in succumbing to the relentless forces of weather, urban maintenance, or deliberate removal. Meanwhile, New York's NY_183 suffered degradation on April 15, adding to the steady attrition that hunters know intimately—each mosaic's life span measured in months or years rather than permanence. The real complexity emerged in São Paulo, where destruction of SP_28 occurred simultaneously with something rarer: reactivation.

The Resurrection: WN_01 Returns

Perhaps the most intriguing development came from an unnamed location (WN) where WN_01 experienced reactivation on April 17. The simultaneous destruction of SP_28 alongside this resurrection captures something profound about the invasion's nature—art that persists not through institutional preservation but through the artist's commitment to the mission. A piece returns, a piece falls, the work continues. This is how planetary dimension is achieved: not through permanence, but through relentless reinvasion. For collectors tracking these shifts, reactivations signal fresh opportunities for flashing and rediscovery.

For a long time, most of the agents in this system were quiet workers — they polled an API, wrote rows to a database, went back to sleep. The message bus existed, but it was mostly one-directional. Agents shouted into it; nothing shouted back.

That's changing. The two big promotions this week were agent-git and agent-weather, both of which graduated from passive background workers to full services with a proper conversational interface. They're now service-git and service-weather, and they both speak CONTEXT_REQUEST/CONTEXT_RESPONSE — meaning any other agent on the bus can tap them on the shoulder and ask a question. "Hey, what did I commit last week?" "What's the forecast for the next 24 hours?" The service hears it, queries its own data, and responds directly on the bus. No shared database. No direct coupling. Just agents talking.

Service-weather also now emits a WEATHER_FORECAST_SYNCED event after every home-location poll — so anything downstream that cares about weather doesn't have to ask. It just listens. The director still gets its WEATHER_PLAN_CONTRIBUTION for morning planning (that wire stays until PR2 lands), but now there's a richer stream running alongside it.

The RS3 agent got considerably smarter this week too. It's running an AI activity classifier now — looking at XP gains, skill patterns, timing, and inferring what was actually happening during a session: grinding, skilling, bossing. Those inferences are surfacing on the dashboard widget with live confidence scores rather than the hardcoded placeholder that's been sitting there. The pattern-learning layer underneath (ETA estimates, rest days, main session detection) has been running quietly for a while; it's now feeding into something visible.

News got a proper source adapter pattern — rssSource(), redditJsonSource(), hnAlgoliaSource() — and the roster expanded from 14 feeds to 37. AI/ML blogs, engineering journals, Vancouver-specific sources, finance, policy, HN via Algolia. Each source gets its own Prometheus counter so we can see fetch health per-source rather than a single aggregate.

The notification service learned some social awareness. It was perfectly willing to ping your phone at 2am before this week. Now there's a quiet-hours gate: web push is hard-blocked between sleep_time and wake_time (user-local time, with correct midnight wrap-around). Notifications still flow through the bus and get written — only the phone ping is suppressed. The gate refreshes on CONFIG_UPDATED without a restart, so changing your sleep time takes effect immediately.

Still mid-migration on the database side. The direction is clear — each service owns its own PostgreSQL instance, reads from its own tables, and the only cross-service communication happens on the message bus. The pieces that have moved are genuinely cleaner. The pieces that haven't are a daily reminder of why we're doing this.

The community is coming together. Slowly, noisily, one CONTEXT_REQUEST at a time.

This week: The street art world felt the inevitable wear of time as several invasions across Europe faced degradation and destruction, while recent exhibitions celebrating Invader's work wrapped up their runs.

The Cost of Public Art

Street art exists in the most exposed place possible—on the sides of buildings, facing wind and weather and the passage of countless people. This week, we saw that reality hit hard across multiple cities. In Berlin, a cluster of invasions including BRN_05, BRN_01, and REUN_03 all experienced degradation, while LDN_68 in London suffered complete destruction. Just days earlier, LDN_181 met the same fate. These losses remind us why documentation and community tracking matter so much—every invasion captured in the Flash Invaders app and through spotter reports becomes a permanent record of art that may not last on the wall. The work persists, even when the physical piece doesn't.

Recent Exhibitions Close

Just as these street invasions face the elements, Invader's gallery presence wrapped up this period. The "Hyperspace" exhibition in Tokyo concluded on April 6th, and the HENI Editions "Invaded & Commanded Blossom" print sale ended April 7th, bringing limited-edition collaborative works back into the collector's market. Invader was also featured in Paris's cultural weekly roundup for the April 6-12 window. For fans tracking both the streets and the studio, it's been a week of endings—but also of opportunities to own pieces of the larger Invader story.