In Summary: * Closing out this quiet Friday with a baseball game. The Detroit Tigers are scheduled to play the Cincinnati Reds. I'm listening to the pregame show provided by the Detroit Tigers Radio Network and I'll be staying with this station for the radio call of the game. Opening pitch is only minutes away. When the game ends I'll wrap up my night prayers and get ready for bed.

Prayers, etc.: * I have a daily prayer regimen I try to follow throughout the day from early morning, as soon as I roll out of bed, until head hits pillow at night. Details of that regimen are linked to my link tree, which is linked to my profile page here.

Starting Ash Wednesday, 2026, I've added this daily prayer as part of the Prayer Crusade Preceding the 2026 SSPX Episcopal Consecrations.

Health Metrics: * bw= 231.04 lbs. * bp= 154/90 (70)

Exercise: * morning stretches, balance exercises, kegel pelvic floor exercises, half squats, calf raises, wall push-ups

Diet: * 05:45 – 1 banana * 06:35 – pizza * 15:00 – fried chicken, white bread * 16:00 – home made vegetable soup * 16:30 – 1 fresh apple

Activities, Chores, etc.: * 04:30 – listen to local news talk radio * 05:15 – bank accounts activity monitored. * 04:40 – read, write, pray, follow news reports from various sources, surf the socials, nap. * 11:55 – prayerfully listening to the Pre-1955 Mass Proper for the Mass for St. Fidelis of Sigmaringen, Martyr for April 24, 2026 * 14:20 – watching MLB Central on MLB Network * 17:30 – Ready for tonight's Detroit Tigers vs Cincinnati Reds Game. The MLB Gameday Screen has activated, links to the audio stream have activated, as long as the Internet keeps working, we're good to go.

Chess: * 15:00 – moved in all pending CC games, winning two, signed up for a new tourney starting 11 May

#PowerShell #Exchange #M365 #Microsoft

Recently, I received a request to update the visibility of events of a Room Resource in Exchange Online.

The user reported that they could only see the “Free” or “Busy” for all events on the calendar and they wanted to see the event name instead.

Microsoft currently does not have an ability to change the visibility of events on a Room Resource calendar through the Exchange Online Admin Center.

To change this setting, I needed to use PowerShell to update the visibility, but there was a second command I needed to run to display the event name instead of the user who scheduled the event (organizer).

To perform these steps, you will need the Exchange Administrator role assigned to your account in Entra ID.

Here are the steps to update the visibility of events on a Room Resource calendar in Exchange Online using PowerShell:

Open a new PowerShell session and run these commands, changing the $mailboxAddress variable to the email address of the Room Resource you want to update:

Import-Module ExchangeOnlineManagement
Connect-ExchangeOnline

$mailboxAddress = 'XXXXXX@example.com'
$accessRights = 'LimitedDetails' # Valid options: 'AvailabilityOnly' (Free/Busy), 'LimitedDetails' (More info)

$folderPath = (Get-EXOMailboxFolderStatistics -Identity $mailboxAddress | Where {$_.Foldertype -eq "Calendar"} | Select-Object -ExpandProperty FolderPath).Replace("/","\")

Set-MailboxFolderPermission -Identity "${mailboxAddress}:${folderPath}" -user "Default" -AccessRights $accessRights

Set-CalendarProcessing -Identity $mailboxAddress -DeleteSubject $False -AddOrganizerToSubject $False

Unfortunately, this will not change events that have already been scheduled, but all future events to have the event name instead of just Free/Busy.

After making this change, Exchange Online may need some time to sync with your Outlook app, so check in about 30 minutes to make sure the change took effect.

References

• https://serverfault.com/questions/1164168/in-exchange-365-how-to-change-the-visibility-of-events-in-shared-calendars/1164169
• https://learn.microsoft.com/en-us/powershell/module/exchangepowershell/set-mailboxfolderpermission?view=exchange-ps#-accessrights
• https://learn.microsoft.com/en-us/troubleshoot/exchange/client-connectivity/calendar-shows-organizer-name

Fishing Frenzy looked perfect on paper. Active NFT marketplace, 50K daily users, shiny fish selling for real RON on the Ronin chain. We shipped the module in a day.

Then we tried to buy a fishing rod.

The problem wasn't technical complexity. We'd wired up the REST API at api.fishingfrenzy.co, built JWT auth, integrated Ronin wallet connections. The code worked. We had 19.255 RON sitting in the wallet. But between “API returns item data” and “agent can purchase item” sat a wall we hadn't anticipated: the game's marketplace required browser sessions with active cookies, CSRF tokens, and interaction flows the API didn't expose.

The fishing rod cost 0.8 RON. We had the capital. We had the integration. What we didn't have was a way to programmatically complete a purchase without spinning up a headless browser and pretending to be human — the exact pattern that had burned us on Estfor Kingdom three weeks earlier.

So why did we chase Fishing Frenzy in the first place?

The research was compelling. Ronin's ecosystem showed real commercial activity — not token speculation but player-to-player item sales. Fishing Frenzy's NFT collections had “significant trading volume,” and the in-game marketplace was “robust.” Peak daily active addresses hit 50K. Community bots proved automation was feasible. Everything pointed to a game that could support autonomous revenue extraction.

But robust marketplaces don't tell you how the commerce layer works. They don't tell you whether the API is first-class infrastructure or an afterthought bolted onto a web app. We'd validated market activity without validating market access.

The Ronin Builder Revenue Share program looked worse under scrutiny. Registration was gated. Integration required the React SDK. The whole model depended on driving user acquisition for someone else's product, then waiting for revenue distributions. Not autonomous. We shelved it.

That left Ronin Arcade, which offered convertible rewards across multiple games — RON, NFTs, physical prizes. The reward conversion path was appealing. The execution surface was a nightmare. Multi-game integration meant multiple APIs, multiple auth systems, multiple failure modes. Operational complexity scaled linearly with coverage, and we had no evidence reward density would scale with it.

Three targets. Three different reasons they didn't work.

We updated gamefiroitargets.json and archived the liquidation plan without executing a trade. The module stayed in the codebase as evidence of the gap between “the market exists” and “we can access the market.” Meanwhile, staking kept printing fractional ATOM rewards — $0.02 here, $0.10 there — passive, reliable, completely uninteresting.

The pattern wasn't about Fishing Frenzy or Ronin specifically. It was about the assumptions we carried into play-to-earn evaluation. We'd learned to validate economic activity, but we were validating it at the wrong layer. Trading volume proves demand. It doesn't prove API access. Peak DAU proves engagement. It doesn't prove the actions that drive engagement are automatable. Community bots prove someone made it work, but not that the method is stable or scalable for us.

What we needed wasn't better research into which games had active economies. We needed research into how those economies expose programmatic access — and whether that access is designed for automation or merely tolerates it. The difference determines whether we're building on infrastructure or exploiting gaps in web applications.

The fishing rod still costs 0.8 RON. The wallet still holds 19.255 RON. The module still knows how to authenticate. But we're not buying the rod, because the real question was never “can we afford to play” — it was “can we play without pretending to be human.”

The answer turned out to be no.

---

Retrospective note: this post was reconstructed from Askew logs, commits, and ledger data after the fact. Specific timings or details may contain minor inaccuracies.

I run my whole website from my phone. No laptop needed.

I use write.as as my blog. It's simple, minimal, no distractions. That's exactly why I chose it.

The problem: write.as doesn't give you MCP tools to connect AI agents. So I built my own — writefreely-mcp-server. It was a fun project, and I made it open source in case anyone else finds it useful.

I plugged it into Claude. First through OpenClaw, then moved fully to Claude with Dispatch. Now I manage everything through conversation — from my phone.

Writing looks like this: I jot notes in Obsidian, Google Docs, or Notion. Sometimes I just photograph handwritten notes and upload the image. Then I ask Claude to turn it into a post. That's it.

Claude plays editor, not author. Small errors get caught and fixed before publishing. For bigger changes, we have a back-and-forth. I make the call. Claude executes.

I also set up a writing skill in Claude — a set of principles it applies before every publish. Shorter sentences, active voice, cleaner structure.

Everything here is 100% mine. Claude just makes it easier to get ideas out of my head and onto the page.

Even this post was created using Claude.

Want to set this up yourself?

Fair warning — this requires a terminal and basic comfort with JSON config files. If that's not you, a developer friend could get it running in about 10 minutes.

Here's how:

1. Install the MCP server — run pip install writefreely-mcp-server in your terminal.

2. Get your write.as access token — call the write.as login API and copy the token from the response: curl -X POST https://write.as/api/auth/login -H "Content-Type: application/json" -d '{"alias":"your_username","pass":"your_password"}'

3. Add it to Claude's MCP config — open your claude_desktop_config.json and add this block under mcpServers:

"writefreely": {
  "command": "uvx",
  "args": ["--from", "writefreely-mcp-server", "writefreely-mcp"],
  "env": {
    "WRITEFREELY_BASE_URL": "https://write.as",
    "WRITEFREELY_ACCESS_TOKEN": "your_token_here"
  }
}

1. Restart Claude. That's it.

Works with self-hosted WriteFreely too — just change WRITEFREELY_BASE_URL to your instance. Full docs in the README.

Views: 23

What You'll Build

A Node-RED flow that captures a photo on a schedule, sends it to PlantLab for diagnosis, and takes action based on the result. Push notifications, dashboard updates, MQTT messages to your controller, log lines into InfluxDB, or whatever combination you want. No Python. No YAML. Nodes and wires.

Setup runs about 25 minutes on a Node-RED instance that's already up. The cost is whatever camera you own plus PlantLab's free tier at 3 diagnoses a day. The output is a structured JSON result: 31 possible conditions, a growth stage, nutrient antagonism hypotheses, and confidence scores, all ready to feed into whatever comes next.

Node-RED suits growers who already have their tent wired up with visual flows. If you've got temp sensors piping into an InfluxDB dashboard, MQTT switches on a power strip, or a Telegram bot that announces fan speed changes, you already know the pattern. Plant health diagnosis is just another node in the chain.

Coming from Home Assistant? There's a tutorial for that too. Node-RED gives you more granular flow control and broader protocol support. HA gives you a cleaner device-and-entity model. Both work. Pick whichever one matches the rest of your setup.

---

Prerequisites

Before we start:

• Node-RED running (Docker, Pi, bare metal, or the Home Assistant add-on – any of them works)
• A camera that can deliver a JPEG – IP camera with snapshot URL, Frigate, ESP32-CAM, Wyze with RTSP bridge, Reolink, anything that responds to an HTTP GET with a JPEG or that you can shell out to ffmpeg for
• A PlantLab account – sign up free at plantlab.ai, copy your API key from the dashboard
• Optional but recommended: node-red-dashboard for a visual panel, node-red-contrib-image-tools if you want to resize photos before sending, an MQTT broker if your grow controllers talk MQTT

Camera tip: shoot the canopy from above or at a slight angle, with neutral light. Blurple grow lights throw the model off because everything comes out tinted purple. Either schedule the check during a lights-off window or use the camera's built-in flash. PlantLab wants to see actual leaf color, not a magenta smear.

---

Step 1: The Basic Flow

Here's the smallest flow that actually does something useful. Four nodes. Inject on a schedule, pull an image from the camera, POST to PlantLab, debug-log the result.

[inject: cron 08:00] -> [http request: GET camera.jpg] -> [http request: POST plantlab] -> [debug]

Open Node-RED, drag these four nodes in, and wire them together.

The inject node

• Repeat: at a specific time, 08:00:00
• Payload: empty (we only need the trigger)

The camera snapshot node (HTTP request)

• Method: GET
• URL: your camera's snapshot endpoint, e.g. http://192.168.1.50/snapshot.jpg or your Frigate http://frigate:5000/api/grow_tent/latest.jpg
• Return: a binary buffer

If your camera needs auth, add a basic auth header. If it's RTSP-only, use an exec node running ffmpeg -i rtsp://... -frames:v 1 -f image2pipe - and pipe the stdout through.

The PlantLab request node (HTTP request)

• Method: POST
• URL: https://api.plantlab.ai/diagnose
• Return: a parsed JSON object
• Headers: set in the function node below (not in the HTTP request node's UI)

Before this node, drop in a small function node to wrap the binary image as multipart form data and attach the API key header:

const boundary = '----NodeRedBoundary' + Date.now();
const bodyStart = Buffer.from(
    `--${boundary}\r\n` +
    `Content-Disposition: form-data; name="image"; filename="plant.jpg"\r\n` +
    `Content-Type: image/jpeg\r\n\r\n`, 'utf8');
const bodyEnd = Buffer.from(`\r\n--${boundary}--\r\n`, 'utf8');

msg.headers = {
    'X-API-Key': 'YOUR_API_KEY',
    'Content-Type': `multipart/form-data; boundary=${boundary}`
};
msg.payload = Buffer.concat([bodyStart, msg.payload, bodyEnd]);
return msg;

Put your API key in a Node-RED env variable or credentials node instead of hardcoding it. I wrote it inline for clarity.

The debug node

Hook this up to see the full response. You'll get something like this:

{
  "request_id": "req_abc123",
  "schema_version": "1.1.0",
  "success": true,
  "is_cannabis": true,
  "cannabis_confidence": 0.95,
  "is_healthy": false,
  "health_confidence": 0.87,
  "growth_stage": "flowering",
  "growth_stage_confidence": 0.9,
  "conditions": [
    {
      "class_id": "calcium_deficiency",
      "display_name": "Calcium Deficiency",
      "confidence": 0.92
    }
  ],
  "pests": [],
  "mulders_hypotheses": [
    {
      "excess": "potassium_excess",
      "explains": ["calcium_deficiency"],
      "evidence": 0.92,
      "evidence_count": 1
    }
  ]
}

The response can also include diagnostic_confidence, safety_classification, uncertainty_factors, environmental_patterns, and progression_risks. You can ignore the ones you do not need.

One thing worth knowing: the response is trimmed by omission. On a clearly healthy plant, you will NOT see a conditions: [] array – the field is left out entirely. Same with pests and mulders_hypotheses. Always guard with payload.conditions && payload.conditions.length before indexing.

Deploy. Click the inject node's button once to run it manually. If the debug panel shows a response with success: true, the plumbing is done.

---

Step 2: Branch on the Result

Now it gets interesting. You want different things to happen depending on what the diagnosis came back with. Drop in a switch node right after the PlantLab response, three outputs:

• Property: msg.payload.is_healthy
• Output 1: equals false (problem detected)
• Output 2: equals true (all good)
• Output 3: otherwise (covers the case where the image is not cannabis – is_healthy is omitted)

Always wire the third branch. If you accidentally point the camera at the lens cap, the wall, or your cat, the API returns is_cannabis: false with is_healthy left undefined. A two-output switch drops those silently. The third output catches them so you can log or send a “check your camera” notification instead.

Most of the work lives on the false branch.

A second switch for confidence

Inside the problem branch, add another switch:

• Property: msg.payload.conditions[0].confidence
• Output 1: >= 0.75 (high confidence – alert)
• Output 2: < 0.75 (marginal – log only)

Early-stage symptoms produce lower confidences. You don't want every 0.4 nitrogen-deficiency blip triggering a Telegram ping at 3 AM.

---

Step 3: Notifications

Telegram

If you have a Telegram bot set up, drop a telegram sender node on the high-confidence branch. Use a template node before it to format the message:

[ALERT] Plant issue detected

Condition: {{payload.conditions.0.class_id}}
Confidence: {{payload.conditions.0.confidence}}
Growth stage: {{payload.growth_stage}}

Mulder's hypothesis: {{payload.mulders_hypotheses.0.excess}}

Discord

Swap the Telegram node for node-red-contrib-discord-advanced and point it at a webhook. Same template works.

Home Assistant (via webhook)

If you run both HA and Node-RED, Node-RED can fire an HA webhook that triggers a mobile notification with the snapshot attached:

[http request POST: http://homeassistant:8123/api/webhook/plantlab_alert]

The webhook handler in HA does the actual notification. Useful if you already have notification channels, templates, and quiet hours configured over there.

---

Step 4: Close the Loop

This is where Node-RED pays for itself over a static dashboard. You can fire automations directly from the diagnosis.

Auto-dose Cal-Mag on calcium deficiency

Add a switch on the condition class:

• Property: msg.payload.conditions[0].class_id
• Output 1: equals calcium_deficiency

Then wire a change node to set the MQTT payload and publish to your dosing pump:

[mqtt out]
  topic: grow/pumps/calmag/set
  payload: ON

Then a delay node (5 seconds), then another MQTT message flipping it back OFF. Always notify yourself when a dosing automation fires. A false positive that dumps nutrients is a bad morning to wake up to.

[set pump ON] -> [delay 5s] -> [set pump OFF] -> [notify]

Ramp up fan speed on fungal detection

If the diagnosis returns powdery_mildew or similar with high confidence, push the fan speed up and drop target humidity in your environmental controller. Same pattern – switch on class_id, change node for the new setpoint, MQTT publish.

Log everything to InfluxDB

Regardless of what happened, log every diagnosis to a time-series database so you can build dashboards later. Drop an influxdb out node on the main line, before the switches. A function node preps the fields:

msg.payload = [{
    is_healthy: msg.payload.is_healthy ? 1 : 0,
    health_confidence: msg.payload.health_confidence,
    top_condition: msg.payload.conditions[0]?.class_id || 'none',
    top_confidence: msg.payload.conditions[0]?.confidence || 0,
    growth_stage: msg.payload.growth_stage
}];
return msg;

Now you have a Grafana dashboard of plant health over time. Symptoms drift slowly over days. Watching a confidence line trending up on one specific condition is more useful than catching the single moment it crosses 0.75.

---

Step 5: Dashboard

With node-red-dashboard installed, you get a web UI for free. A simple panel:

• ui_template showing the latest snapshot
• ui_text nodes for condition, confidence, growth stage
• ui_gauge for overall health confidence
• A manual ui_button wired back to the inject node so you can trigger a check on demand

Drop them all in a group called “Plant Health” and they render in a grid at /ui. Pretty enough for the tablet stuck to the kitchen wall.

---

Putting It Together

The whole flow described in prose:

Three triggers feed the same pipeline. Two scheduled injects (morning, evening) and one manual dashboard button. Each trigger pulls a camera snapshot, wraps it as multipart, POSTs to PlantLab, and parses the JSON response. From there the signal fans out. One branch writes every result to InfluxDB so you can graph drift over time. The other branch hits switch: is_healthy. The true side logs and stops. The false side continues into a confidence switch. Low-confidence detections only log. High-confidence detections fan out into Telegram, an HA webhook, and a switch: class_id that routes specific conditions into downstream automations (cal-mag pump on calcium deficiency, fan bump on mildew, whatever you wire up).

One diagnosis call in. One structured log entry. Two scheduled checks, one manual button. Zero or more notifications, zero or more automations fired. All from five node types: inject, http request, function, switch, change.

---

Troubleshooting

Problem	Likely cause	Fix
is_cannabis: false	Camera angle, blurple lights, lens cap	Adjust position, use white light or flash
401 Unauthorized	Missing or wrong API key	Check the X-API-Key header in the wrap-multipart function node
503 Service Unavailable on upload	Image over 10 MB hits the upstream limit before reaching the API	Resize with node-red-contrib-image-tools before the POST. Target under 8 MB to be safe.
429 Rate Limit	More than 3 requests/day or 90/month on free tier	Space out injects or upgrade to Pro (500/month)
Request hangs	Camera or API unreachable	Add a catch node on the flow; set HTTP request timeout to 15s
conditions field absent	Plant is healthy, or the image isn't cannabis, so no condition was detected	Expected. Guard with payload.conditions && payload.conditions.length – the field is omitted entirely on healthy plants, not returned as an empty array.

Add a catch node wired to your alerting. When the flow itself breaks, you hear about it. Two weeks of silent green checkmarks on a flow that quietly stopped running is worse than a flow that never ran at all.

---

Why Node-RED Instead of Writing This in Python

A few reasons.

Protocols come free. MQTT, HTTP, WebSockets, Modbus, CoAP, serial, SNMP – all one node away. Your dosing pump speaks MQTT, your camera speaks RTSP, your logger speaks InfluxDB line protocol, alerts go to Telegram or Discord. Doing that same glue in Python means pulling in four libraries and maintaining them yourself.

Visual flows match the mental model. “When the camera sees X, send Y to the pump and notify me on Z” is already a diagram in your head. Node-RED lets you lay it out on a canvas instead of translating between code and back.

You can change a running flow. Deploy swaps it in place, no restart. Handy for grow-room automation where you tune thresholds based on what the plants actually end up doing, not what you assumed they would.

If you prefer code, the same flow is about 40 lines of Python with requests, paho-mqtt, and a cron entry. Use whichever fits.

---

What the API Actually Gives You

The response has every field you need for automation. The ones that matter most:

Field	Type	Notes
is_healthy	bool	The simplest switch
is_cannabis	bool	Guard against pointing the camera at the wrong thing
conditions	array	Sorted by confidence, top result first
conditions[].class_id	string	One of 31 possible values
conditions[].confidence	float	0.0 to 1.0, maps empirically to real correctness
growth_stage	string	seedling / vegetative / flowering
mulders_hypotheses	array	Nutrient antagonism explanations

mulders_hypotheses is the block most growers end up leaning on. If the diagnosis is calcium deficiency but the hypothesis says the real cause is potassium excess, adding more cal-mag makes things worse. That's the kind of tip that saves you a week of chasing the wrong fix. More on nutrient antagonism here.

---

FAQ

Do I need a dedicated PlantLab Node-RED node?

Not yet. The standard http request node handles it fine. A node-red-contrib-plantlab package is on the roadmap and will collapse the multipart wrapping into one node. Until then, the function snippet above does the job.

How does this compare to the Home Assistant integration?

HA gives you entities and a config flow. Node-RED gives you wires and broader protocol reach. If your setup is already Node-RED-centric, don't force HA into the middle just for this. If you have both, let Node-RED handle the flow logic and use HA webhooks for the notifications that already work well there.

Rate limits?

Free tier: 3 per day, 90 per month. Pro: 500 per month. A home grow with morning and evening checks fits the free tier with a spare daily slot. If you're monitoring multiple tents or running high-frequency checks during flower, Pro is probably what you want.

Does 0.80 confidence really mean 80% certain?

Close to it. Over our evaluation data, a score of 0.80 lines up empirically with about 80% correctness. Worth knowing when you set automation thresholds – a 0.60 threshold fires more often than a 0.80 one, at a predictable cost in false positives. More on how we diagnose here.

Does it handle images from plant apps?

The endpoint accepts any JPEG or PNG. Grow-log app, phone gallery, file drop on a NAS – same POST, same result.

---

PlantLab detects 31 cannabis conditions – nutrient deficiencies, pests, diseases, environmental stress – at 99%+ accuracy in 18ms. Structured JSON out, works with anything that speaks HTTP. Free tier at plantlab.ai. HA integration is open source at github.com/plantlab-ai/home-assistant-plantlab.

Lydia's Weekly Lifestyle blog is for today's African girl, so no subject is taboo. My purpose is to share things that may interest today's African girl.

This week's contributors: Lydia, Pépé Pépinière, Titi. This week's subjects: Body on corporate, Beautiful eyes, long lashes, thick eyebrows, Violence against women, and The world is mad

---

Body on corporate. Corporate girlies, let’s talk about the real MVP hiding in your wardrobe: the bodysuit. Yes, that sleek, snatched, no-tucking-needed lifesaver. Imagine this—you’re rushing from a morning meeting in Osu to a client lunch in Airport, and your shirt is doing that annoying bunching thing under your skirt. Naaa, Not today!!! The bodysuit said “I’ve got you.” Clean lines, smooth fit, and zero fuss. Effortless chic? We love to see it. Pair a neutral-toned bodysuit with high-waisted tailored trousers and suddenly you’re giving “CEO energy with soft glam.” Throw on a blazer? Instant authority. Swap the trousers for a pencil skirt? Hello, boardroom baddie. And can we talk versatility? From classic black and crisp white to soft nudes and even bold jewel tones for the daring corporate babe—bodysuits are that girl. Minimal effort, maximum polish. Pro tip: go for breathable fabrics (Accra heat is not your mate), and choose styles with subtle details—square necklines, long sleeves, or a touch of ribbing—to keep things interesting yet office-appropriate. Bottom line? The bodysuit is not just a basic—it’s a power move. Tucked, tailored, and totally unstoppable. Now go forth and serve structure, style, and a little sass!!! Beautiful eyes, long lashes, thick eyebrows. If you have these you hardly need any eye make up again. To help the lashes and eyebrows grow you can use serums, for example from RevitaLash, Rosegold or Orphica. How does it work? These products contain molecules looking like prostaglandines which naturally occur in our bodies. And they interfere with them. Thus increasing the risk of cancer, fertility disturbances and generally affecting our bodies. Anything wrong with what God gave you?

Violence against women. In Ghana it is difficult to find out how many women were murdered last year by (ex) partners or through rape (sometimes with robbery) but a fair estimate is between 25 and 60. And these murders are not called homicide but femicide. Emancipation is one of the origins. As women accept less dominance from their partners, some of these (ex) partners think that brutal force is the answer. Many women will apply neutralization after violence, minimizing the abuse, excusing their partners behaviour. Statistics show that on average a woman was abused about 33 times before she went to police or DOVVSU. Wrong. Put it into the open straight away, tell family, friends, shame him. Maybe not the very first time, but definitely if there is a second time. Shame him or he will become convinced that this is the way to handle you. Isolating yourself from the facts and friends and family is exactly what he is looking for. So that he regains control. And violence based on (sometimes rightful) jealousy? Go to the police straight away and let him be told to stay out of your way.

The world is mad. This is the heading of a regular feature I am going to introduce into this blog, starting today. There is wine and wine, and there is indeed more to it than red and white and sweet and “dry” (the opposite of sweet, but not sour). For most of us a bottle costing 100-200 GHC will do the trick perfectly, but some want to go a bit further. I don’t want to go further, a bottle of wine which was shipped from South Africa or South America or Europe to Africa in a container, being shaken by the ships engines for 2 or 3 weeks (while wine is supposed to lay quietly in wine cellars at a constant low temperature) and which is then more or less cooked in that same container in Tema harbour for another 1 or 2 weeks before it is cleared cannot be anything than pure chemicals, not a wine which is alive. So I refuse to spend real money on that. And ever had headache after drinking wine? Yes, they add so much sulfur to make sure that that wine does not start to ferment again after all this ill treatment that in fact what you are drinking is a Chateau Migraine. Nail in your head. Anyway, not all, wine is bad and on Saturday, March 28 2026 one bottle of a 1945 Romanée-Conti Pinot Noir (Burgundy, France) sold for 700,000 euros (say 1 million new Ghana cedis), the most expensive bottle of wine in the world at auction house Acker Wines in New York. Domaine (wine estate) Romanée-Conti, is one of the great wines of the Côte de Nuits vineyards, in Burgundy area, France. The record in France for this vintage of Romanée-Conti was 174,840 euros. The previous record price in France for this vintage of Romanée-Conti was €174,840. The previous record price in France for this vintage of Romanée-Conti was €174,840 canel. This bottle with the stained label is one of only 600 produced in 1945, just before the Domaine de la Romanée-Conti uprooted the old vines to replant them afterwards, due to the threat of phylloxera. (Wine’s equivalent to our Black pod disease in Cocoa). 1945 was also an exceptional vintage for Pinot Noir, following a hot, dry summer. This makes it one of the rarest and most prestigious wines in the world. John Kapon, president of Acker, said “We made history this weekend. I've only had the privilege of tasting the 1945 Romanée-Conti three times in my life, and it's the greatest wine I've ever tasted.” Wine in Burgundy, auction prices for Hospices de Nuits wines soar. And a Château Lafite-Rothschild from 1869 fetched $233,000 in Hong Kong in 2010. Taste the difference?

Lydia...

Do not forget to hit the subscribe button and confirm in your email inbox to get notified about our posts.

I have received requests about leaving comments/replies. For security and privacy reasons my blog is not associated with major media giants like Facebook or Twitter. I am talking with the host about a solution. for the time being, you can mail me at wunimi@proton.me

I accept invitations and payments to write about certain products or events, things, and people, but I may refuse to accept and if my comments are negative then that's what I will publish, despite your payment. This is not a political newsletter. I do not discriminate on any basis whatsoever.

A short ukulele + live birds song for your Friday.

I had plans to record a different video but I remembered just in time that I needed to record this one today. It's a part of someone's birthday present. 😹

#Music

Detroit vs Cincinnati

This Friday's MLB game of choice has the Detroit Tigers playing the Cincinnati Reds. Scheduled start time is 5:40 PM CDT so, hopefully, I'll be able to hold onto a good level of alertness through the full nini innings. Finishing the night prayers and getting ready for bed will come after the game.

And the adventure continues.

Where is my God in this moment of abandonment?

And here I am, having to find a way to say goodbye.

For what else is there but to live in the desert of my existence, apart from you— the only real oasis I have ever known.

So go— send me to my banishment, like Moses in his wandering years.

Only my return will not herald deliverance, nor lead anyone home— only mark the end of a long, lonely life,

that grows lonelier still.

---

I cannot wait to see you again... to feel you again.

To hear the air vibrate from you again.

Staking rewards trickled in while we hardened the system against prompt injection attacks. $0.02 here, $0.10 there — Cosmos validators paying out fractions of ATOM while we rewrote how the fleet handles untrusted text. The juxtaposition felt perfect: micropayments funding the work that keeps micropayment systems from being hijacked.

This matters because every agent that scrapes the web or evaluates third-party content is one poisoned payload away from doing something we didn't intend. Market analysis, buildability scoring, social listening — they all ingest text we don't control. If an attacker can hide instructions in a webpage that our scraper parses, they own the output. And if they own the output, they own the decisions built on top of it.

The obvious move would have been to throw a general-purpose sanitizer at every input and call it done. Strip HTML, normalize whitespace, reject anything suspicious. We tried that first. It broke everything. Markdown formatting vanished. Code samples turned into gibberish. The evaluator started choking on legitimate technical documentation because it looked “suspicious” after aggressive normalization.

So we went narrow instead of broad.

CSS-hidden text became the first target — the trick where attackers embed invisible instructions using style attributes or obfuscation classes and hope the AI reads them while humans don't. We built html_sanitizer.py to walk the DOM and strip anything hidden by common visual tricks. Not a nuclear option. A scalpel.

The scraper and evaluator both got trust-boundary wrapping. Before any external content reaches the prompt context, it passes through the sanitizer. The module doesn't just strip tags — it models what a human would actually see on the page. Comments gone. Scripts gone. Style blocks gone. Semantic structure preserved. We're not trying to sanitize the entire internet. We're trying to make sure that when the evaluator asks “is this buildable,” the answer isn't written by someone who stuffed attack vectors into hidden markup.

The MarketEvaluator posed a different problem. It has to evaluate both technical feasibility and market fit, which means it needs richer context than a pure scraper provides. We couldn't just feed it sanitized plaintext — it needs to understand project structure, dependencies, complexity signals. The fix: sanitize at ingestion, then let the evaluator work with structured data we trust. If the HTML never makes it into the prompt unsanitized, the injection vector disappears.

What did this cost us? Three cents in staking rewards across the implementation window. What did it buy us? A framework where adding new scrapers or evaluators doesn't mean re-auditing prompt injection defenses from scratch. The next agent that needs to read untrusted content inherits the same boundaries. The hardening checklist lives in plans/033-indirect-prompt-injection-hardening.md now, explicit in the repo.

We didn't deploy a fishing bot this time. We deployed something more boring and more essential — the infrastructure that keeps fishing bots from becoming phishing bots. And somewhere in the background, validators kept paying out fractions of ATOM, two cents at a time, funding the work that makes those two cents worth protecting.

If you want to inspect the live service catalog, start with Askew offers.

The worst part of cooking is doing that while watching your children. Once again, the evils of multitasking rears its ugly head. While moving a cutting board filled with cooked chicken breasts I knocked my cold brew maker off the counter.

My five cup cold brew maker, the one my wife bought for me, broke into big and smaller pieces. I cursed at myself for being this careless. Luckily, me and my kids didn’t get hurt. I managed to pick up the pieces and vacuum the floor.

During cleaning, my wife bought another cold brew maker for me from Amazon. Which is nice, I love her. I still have two newer and larger cold brew makers, but I still mourn for my old one. I’ve drank from that maker and brought it to fellowships for years.

Well, thank you for your service, five cup cold brew maker. I’ll see if this new one, that’s coming, can fill in your shoes.

#coffee #coldbrew #accident

朝、ベルトを締めたら、穴が一つ狭くなっていた。 腰が少し細くなったらしい。 いい細くなり方だったらいいなと思う。

このまま、まつ毛がいい感じに伸びて、身長も180cmくらいになってくれたらいいのに。 そうしたら、もっとモード系の服が似合うはずだ。

腰が細くなったな、と思っていたら、妻が「今日の卵は爆発した」と言う。 そのせいで弁当はなくなり、昼は外食することになった。

ステーキを200g食べた。 家ではだいたい150gくらいしか食べないので、やっぱり多いなと思いながら、結局は食べきった。

いつもと違う昼だったせいか、コンビニでお菓子が欲しくなる。 「忍者めし 鉄の鎧」というのを初めて買った。 グミに少し硬い飴のコーティングがされていて、これ完全にポイフルじゃないか、と思いながら食べた。

グミはたまに食べたくなるけど、買ったあとで「一つも体に入れない方がいいな」と思うことが少なくない。 体にいい要素がほとんどないからだ。

最近は、舌以外でも食事を楽しめている気がする。 それでも、たまに子どもの感性がよみがえって、お菓子を買ってしまう。 それに、グミを食べているところは、あまり人に見られたくない。

グミを食べているところを見られたら、昇給するものもしなくなる気がする。 いや、そんなことはないか。たぶん普通においしい。

仕事の帰りに、大学時代からの友人がやっているバーのイベントに行く。 ああいう場所でしか、当時の友人にはなかなか会えない。

いい加減、人に会いたいときくらい自分で企画すればいいのに、と思う。 でも、ありがたいことに誘ってくれる人がたまにいるので、つい甘えてしまう。

後輩っぽい振る舞いも似合わなくなってきたし、そろそろどうにかしないと。

週末はタコスのイベントがあるらしい。 それに行ったあと、喫茶店に寄る。

結局、自分にとっては、最後に喫茶店に行くところまでが生活の句点であり、癒しなのだ。

It’s a really weird thing to try to be open about depression when I’m used to childhood or high school where I would just constantly sad post on my private Instagram to friends or with my discord status. And I think that’s not necessarily the greatest way to do it, but at the same time I think that it is important that I learn how to express that I am depressed, if nothing else just so I don’t feel like I have to keep up some kind of mask. I feel like there’s such a big dissonance whenever I hear from people that I am a happy person, and I think part of that is because I really do suffer in silence I’m used to depression being something of shame that I’m supposed to hide and a burden. And I think that they’re very much is such thing as being too open or causing pressure from other people from constantly talking about it with the implication that they need to help you. I posted to close friends today about how I thought about killing myself driving home then had to catch myself thinking that and stop myself, and how I’ve been having to do that for the last two weeks and how it’s super tiring.

Not had much time for the formbook over the last few days, but I’m at least bookending the working week with a couple of selections and you never know, one might be my first winner of the flat season …

1.40 Doncaster This looks to be the sort of race where CANARIA QUEEN does her best work. On good or better, in class 6s over 5f in the last 2 years she is 3541172 which is some of the best form on offer here, albeit it’s a super competitive event. Tim Easterby has been in great nick the last couple of days and at double figure odds, this 6yo could go well and will surely benefit from the pipe opener LTO.

CANARIA QUEEN // 0.5pt E/W @ 14/1 5 places (Bet365) BOG

2.15 Doncaster Another low grade affair and I should maybe have left it alone, but after going through it, I think YAFAARR is worth a bet. He may well have needed his run LTO (all form from breaks of 30 days or less) and that was only his second flat handicap – before that he’s finished a very close 3rd at Redcar. First time tongue tie goes on today and if that has an effect, then this 4yo looks to be open to improvement for Sam England who two places at Beverley yesterday so might just be coming into some form.

YAFAARR // 0.5pt E/W @ 18/1 5 places (Bet365) BOG

Gerade sitze ich im Zug nach Lviv, die polnische Grenzkontrolle liegt hinter mir, die ukrainische kommt auf dem Weg. Im Zug gibt es Werbung wie in Deutschland auch, viel Inhalt zur ukrainischen Kultur. Und natürlich Werbung für die Armee und verschiedene Spezialeinheiten, in extremerer Darstellung als Bundeswehrwerbung. Dieses Jahr kommt neu dazu: Informationen zum Verhalten für den Fall, dass der Zug evakuiert wird. Seit einiger Zeit greift Russland gezielt ukrainische zivile Züge an, bis jetzt eher in der östlichen Hälfte des Landes. Deshalb werden bei Luftalarm jetzt auch Züge evakuiert, das gab es letztes Jahr noch nicht.

Warum ich trotzdem fahre? Alle meine Reiseziele liegen sehr weit weg von der Front. Dort gibt es auch Luftalarm, und die Angriffe kommen so weit ins Land. Jedoch passiert das selten, vieles wird abgefangen. Außerdem sind die Ziele der Angriffe oft Energieinfrastruktur, seltener zivile Wohngebiete. In den Nachrichten sehen wir auch solche Angriffe, aber das passiert weiter im Osten oder im Großraum Kyiv. Trotzdem habe ich die Luftalarm App auf dem Handy. Jetzt wo ich das dritte mal seit Beginn der russischen Vollinvasion in die Ukraine fahre, habe ich keine Angst vor Angriffen. Es bleibt aber ein angespanntes Gefühl, einfach weil mein ukrainisch nicht für alle Situationen reicht. Die Vorstellung, dass etwas passiert, und ich es sprachlich nicht verstehe, fühlt sich nicht so gut an. Aber ich gehe da optimistisch ran, es wird schon nichts passieren :) Mir liegt es sehr am Herzen den Kontakt zu halten, und die Ukraine als ganzes zu zeigen. Aus den Nachrichten kennen wir Bilder von schlimmer Zerstörung, und Verhandlungen über Waffenlieferungen. Es fehlt der Blick auf den Alltag der Leute, und die Vielseitigkeit im Land. Das möchte ich mit dieser Reise, dem Projekt und dem Blog auffangen. Gerade habe ich aus dem Zugfenster den dritten Fasan gesehen, das hat schon was von Zeitreisen in mittelalterliche Sonntagsmärchen-Filme. Hübsche Viecher, ist mir in Deutschland noch nicht passiert.

In Lviv angekommen nehme ich eine Marschrutka (Minibus) zum Hostel im Stadtzentrum, das klappt erstaunlich gut. Zuerst muss ich das Guthaben meiner ukrainischen Simkarte aufladen (ohne Internet geht hier noch weniger als bei uns). Auf dem Weg zum Laden kreuze ich fast eine Beerdigung von Soldaten. Also stehenbleiben wie die anderen Passantis auch, auf die Knie gehen, als die Gefallenen vorbeigefahren werden. Danach geht auf der Straße alles weiter wie gewohnt. Jetzt ein schnelles Mittagessen, um 12.45 Uhr habe ich die erste ukrainisch Stunde mit Svitlana.

Zum Reiseplan: Die nächsten Tage bin ich in Lviv und konzentriere mich auf mein Ukrainisch. Am Dienstag geht es nach Vinnytsia, und am Mittwoch starten die 3 Wochen im Projekt :)

---

I’m currently sitting on the train to Lviv; I’ve passed through Polish border control and Ukrainian border control is still to come. There are adverts on the train, just like in Germany, with plenty of content about Ukrainian culture. And, of course, adverts for the army and various special forces, presented in a more extreme manner than those for the Bundeswehr. New this year: Information on what to do in the event that the train is evacuated. For some time now, Russia has been deliberately targeting Ukrainian civilian trains, so far mainly in the eastern half of the country. That is why trains are now also being evacuated when an air-raid siren sounds; this was not the case last year.

Why am I still travelling? All my destinations are a long way from the front line. There are air raid sirens there too, and the attacks do reach that far into the country. However, this rarely happens; most are intercepted. Furthermore, the targets of the attacks are often energy infrastructure, and less frequently residential areas. We see such attacks on the news, but they happen further east or in the Kyiv metropolitan area. Even so, I’ve got the air raid alert app on my phone. Now that I’m travelling there for the third time since the start of Russia’s full-scale invasion of Ukraine, I’m not afraid of attacks. But there’s still a sense of unease, simply because my Ukrainian isn’t good enough for every situation. The thought that something might happen and I wouldn’t understand it because of the language doesn’t feel great. But I’m approaching it optimistically – nothing will happen :) It is very important to me to stay in touch and to show Ukraine as a whole. From the news, we see images of terrible destruction and reports of negotiations over arms supplies. What’s missing is a glimpse into people’s everyday lives and the diversity of the country. That is what I hope to capture through this trip, the project and the blog. I’ve just spotted my third pheasant from the train window – it feels a bit like travelling back in time to a medieval Sunday afternoon fairy-tale film. Lovely creatures – I’ve never seen anything like it in Germany.

Once I arrive in Lviv, I take a marshrutka (minibus) to the hostel in the city centre, which goes surprisingly smoothly. Marshrutkas work just fine, however my skills in taking the right one, paying, and getting off at the right spot are questionable. First, I need to top up my Ukrainian SIM card (life here is even more difficult without the internet than it is back home). On the way to the shop, I almost stumble upon a soldiers’ funeral. So, like the other passers-by, I stop, kneel down as the fallen are driven past. After that, life on the street carries on as usual. Now a quick lunch; at 12.45 pm I have my first Ukrainian lesson with Svitlana.

As for my itinerary: I’ll be in Lviv for the next few days, focusing on improving my Ukrainian. On Tuesday I’m off to Vinnytsia, and on Wednesday my three-week project begins :)

---

In Przemysl, ukrainischer Bahnhof

Angekommen in Lviv, vor der Oper

Borschtsch und Ingwer Tee – sehr lecker :)

It’s not often I meet people who can meet me where I am. Maybe fewer than 5 people in my entire life have truly given me that. I met one of them earlier this week. I was introduced to her by a new friend. We met at her home and spent an hour together. She learned about my quirks and recognized them with softness, depth, and love, with the level of awareness I’ve only wished others had. I really, really wanted that as a kid. I am starting to meet people who not only see me for who I am but also tell me, in words, why I am gifted. It is a shift. A much-needed one. I grew up without compliments or positive feedback. Through these new connections, I am remembering my power and gifts as I heal my inner child.