[JAlcocerTech] Services and Execution

Tl;DR

Still thinking on headcounts to mess around with a project instead of getting shit done outcomes?

Intro

• WHY Im writting this post: bc I wanted to continue the Home x IoT Improvements, made the mbsd 0-7-0 release and used cloudflare KV to get leads
• What Ive learnt with it: Ive ended up telling agents the WHY, not the how, via PIO fwk

A friend told me once that I will do sth with energy at some point

Another, the coding is my thing

It seems that both were right.

Updates

Since the last one:

flowchart LR
    %% --- Styles ---
    classDef free fill:#E8F5E9,stroke:#2E7D32,stroke-width:2px,color:#1B5E20;
    classDef low fill:#FFF9C4,stroke:#FBC02D,stroke-width:2px,color:#FBC02D;
    classDef mid fill:#FFE0B2,stroke:#F57C00,stroke-width:2px,color:#F57C00;
    classDef high fill:#FFCDD2,stroke:#C62828,stroke-width:2px,color:#C62828;
    classDef bridge fill:#E3F2FD,stroke:#1565C0,stroke-width:3px,color:#0D47A1;

    %% --- Nodes ---
    L0("Free Content
( DIY = $0)"):::free
    L1("Web Audits 🛡️
(Reveals Problem )"):::free
    L11("Tech Blog/Youtube"):::free
    L12("ebooks"):::free
    L13("mbsd framework OSS"):::free
    L14("OSS guides"):::free

    L3("Done With You
(Trade $$ for knowledge)"):::mid
    L4("Done For You
(Trade $$$ for outcomes)"):::high
    L44("GenBI
Shopify PoC"):::bridge
    L45("Real Estate
Funnel Bot"):::bridge
    L46("Energy Solutions
HVAC"):::bridge
    L47("IoT Solutions
Crops"):::bridge
    L48("Weddings
Photo QR"):::bridge

    %% --- Connections ---
    L0 --> L1
    L1 --> L3
    L12 --> L3
    L13 -->|MultiBodySystemsDynamicscom| L3
    L14 -->|FOSS Engineer| L3
    L0 --> L11
    L0 --> L12
    L0 --> L13
    L0 --> L14
    L3 --> L4
    L4 -->|Productized Service| L44
    L4 -->|Productized Service| L45
    L4 -->|Productized Service| L46
    L4 -->|Productized Service| L47
    L4 -->|Productized Service| L48

MBSD

Its been few weekly releases for the multi body OSS framework:

• https://github.com/JAlcocerT/mbsd-core
• https://ebooks.jalcocertech.com/books/mechanism-analytics/

All linked to: https://multibodysystemsdynamics.com/ for which I have the web UI repo here.

scp jalcocert@192.168.1.2:/home/jalcocert/multibody-tests/v-0-6-0-concerns.md . 
#scp jalcocert@192.168.1.2:/home/jalcocert/multibody-tests/whitepaper.md . 
#scp jalcocert@192.168.1.2:/home/jalcocert/multibody-tests/roadmap.md .

#claude --dangerously-skip-permissions -p "promptwhateverrrr" #yolo
cd /home/jalcocert/Desktop/mbsd-framework/mbsd-core

git switch main
git merge --ff-only v0.7.0-dev
git tag -a v0.7.0 -m "MBSD Core v0.7.0"
git push origin main
git push origin v0.7.0

awk '
  /^## v0\.7\.0 / { found=1; next }
  /^## / && found { exit }
  found { print }
' CHANGELOG.md | gh release create v0.7.0 \
  --repo JAlcocerT/mbsd-core \
  --verify-tag \
  --title "MBSD Core v0.7.0 - Spatial Kinematics Preview" \
  --notes-file - \
  --latest

cd /home/jalcocert/Desktop/mbsd-framework/mbsd-examples

git switch main
git merge --ff-only v0.7.0-dev
git tag -a v0.7.0 -m "MBSD Examples v0.7.0"
git push origin main
git push origin v0.7.0

awk '
  /^## v0\.7\.0 / { found=1; next }
  /^## / && found { exit }
  found { print }
' CHANGELOG.md | gh release create v0.7.0 \
  --repo JAlcocerT/mbsd-examples \
  --verify-tag \
  --title "MBSD Examples v0.6.0 - Spatial Kinematics Preview" \
  --notes-file - \
  --latest

cd /home/jalcocert/Desktop/mbsd-framework/mbsd-core

uv build
uvx twine check dist/*
uvx twine upload dist/mbsd-0.7.0*

Users could then install with:

pip install mbsd==0.7.0

cd ~/Desktop/mbsd-framework/multi-body
git switch -c web-sleek-and-app

git submodule add https://github.com/JAlcocerT/mbsd-core.git external/mbsd-core
git submodule add https://github.com/JAlcocerT/mbsd-examples.git external/mbsd-examples
git -C external/mbsd-core checkout v0.7.0
git -C external/mbsd-examples checkout v0.7.0
git add .gitmodules external/mbsd-core external/mbsd-examples
git commit -m "Reference mbsd-core and mbsd-examples v0.7.0 as submodules"

#git subtree add needs a clean working tree, so do this before building anything. ./web is left
#untouched.

git subtree add --prefix=web-sleek ../multibodysystemsdynamics-site main
git subtree add --prefix=app       ../app-mbsd main

mkdir -p app/vendor
(cd external/mbsd-core && uv build --wheel --out-dir ../../app/vendor)
ls -la app/vendor # expect mbsd-0.7.0-py3-none-any.whl (~54 KB)
sha256sum app/vendor/mbsd-0.7.0-py3-none-any.whl

# # app: use the committed wheel unless MBSD_WHEEL is overridden
# sed -i '/^CF_PAGES_BRANCH/a export MBSD_WHEEL ?= $(CURDIR)/vendor/mbsd-0.7.0-py3-none-any.whl' app/Makefile
# # app: fallback core checkout + CPython demo tests now live in external/
# sed -i 's|"\.\./mbsd-core"|"../external/mbsd-core"|' app/scripts/sync-mbsd-wheel.mjs
# sed -i 's|"test:demos": "../mbsd-core/.venv/bin/python tests/test_demos.py"|"test:demos": "uv run --project ../external/mbsd-core python tests/test_demos.py"|' app/package.json
# # web-sleek: the snippet check runs against external/mbsd-core
# sed -i 's|^MBSD_PYTHON ?= ../mbsd-core/.venv/bin/python|MBSD_PYTHON ?=../external/mbsd-core/.venv/bin/python|' web-sleek/Makefile
# sed -i 's|ROOT.parent / "mbsd-core"|ROOT.parent / "external" / "mbsd-core"|'web-sleek/scripts/verify-snippets.py
# 1. Remove the unwanted .gitignore
rm app/vendor/.gitignore

# 2. Fix path in verify-snippets.py
sed -i 's|ROOT.parent / "mbsd-core"|ROOT.parent / "external" / "mbsd-core"|' web-sleek/scripts/verify-snippets.py

# 3. Clean up spacing in web-sleek/Makefile
sed -i 's|^MBSD_PYTHON ?=\.\./|MBSD_PYTHON ?= ../|' web-sleek/Makefile

# Confirm settings
git check-ignore -v app/vendor/mbsd-0.7.0-py3-none-any.whl || echo "wheel not ignored: ok"
grep -n 'external' web-sleek/scripts/verify-snippets.py

# Sync and run test suites
(cd external/mbsd-core && uv sync)
(cd app && npm ci && make test && make build)
(cd web-sleek && npm ci && make check)

# Check status, stage, commit, and push
git status --short
git add app/vendor/mbsd-0.7.0-py3-none-any.whl \
        app/Makefile \
        app/package.json \
        app/scripts/sync-mbsd-wheel.mjs \
        web-sleek/Makefile \
        web-sleek/scripts/verify-snippets.py
git commit -m "Vendor mbsd 0.7.0 wheel for the app; point web-sleek and app at external/"
git push -u origin web-sleek-and-app

gh pr create --base master --head web-sleek-and-app --fill
gh pr merge web-sleek-and-app --merge --delete-branch
git switch master && git pull
git submodule update --init external/mbsd-core external/mbsd-examples

# Define your legal metadata
ENTITY="Your Name or Company S.L."
ADDRESS="Street 1, 28001 Madrid, Spain"
AUTHORITY="Agencia Española de Protección de Datos (aepd.es)"
JURISDICTION="Spain"

# Replace placeholders across legal markdown files
sed -i -e "s|{{LEGAL_ENTITY}}|$ENTITY|g" \
       -e "s|{{LEGAL_ADDRESS}}|$ADDRESS|g" \
       -e "s|{{SUPERVISORY_AUTHORITY}}|$AUTHORITY|g" \
       -e "s|{{JURISDICTION}}|$JURISDICTION|g" \
       web-sleek/src/content/legal/*.md

# Verify and commit
(cd web-sleek && make check-legal)
git commit -am "Fill in legal details" && git push

# Authenticate and inspect account
npx wrangler login
npx wrangler whoami
npx wrangler pages project list

# Create production projects
npx wrangler pages project create app-mbsd --production-branch main
npx wrangler pages project create multibodysystemsdynamics-site --production-branch main

# Create KV storage for lead collection (copy the printed ID)
npx wrangler kv namespace create LEADS

# (Optional) Add notification webhook secret
npx wrangler pages secret put LEAD_WEBHOOK_URL --project-name multibodysystemsdynamics-site

# Commit KV configuration
git commit -am "Bind LEADS KV namespace" && git push

# Deploy app and site
(cd app && make deploy)
(cd web-sleek && make deploy)

# Test contact form API submission (expects HTTP 202)
curl -s -o /dev/null -w "%{http_code}\n" -X POST \
  https://multibodysystemsdynamics-site.pages.dev/api/contact \
  -H "Content-Type: application/json" \
  -d '{"name":"Deploy test","email":"you@example.com","project":"Something else","details":"test"}'

# Verify lead storage in KV (replace <LEADS_ID>)
npx wrangler kv key list --namespace-id <LEADS_ID> --remote

# Set your Cloudflare variables
ACCOUNT="<account-id>"
TOKEN="<api-token>"
OLD="<v1-project-name>"
API="https://api.cloudflare.com/client/v4/accounts/$ACCOUNT/pages/projects"

# 1. Attach app subdomain (zero downtime)
curl -s -X POST "$API/app-mbsd/domains" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"name":"app.multibodysystemsdynamics.com"}'

# 2. Reassign apex domain from v1 to new site
curl -s -X DELETE "$API/$OLD/domains/multibodysystemsdynamics.com" \
  -H "Authorization: Bearer $TOKEN"

curl -s -X POST "$API/multibodysystemsdynamics-site/domains" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"name":"multibodysystemsdynamics.com"}'

# 3. Check status until state shows "active"
curl -s "$API/multibodysystemsdynamics-site/domains" \
  -H "Authorization: Bearer $TOKEN" | python3 -m json.tool | grep -E '"name"|"status"'

for p in / /tools/ /glossary/grashof-condition/ /case-studies/practica/ /legal/privacy/ /rss.xml; do
  curl -s -o /dev/null -w "%{http_code} $p\n" "https://multibodysystemsdynamics.com$p"
done
curl -s -o /dev/null -w "%{http_code} app\n" https://app.multibodysystemsdynamics.com/

cat >> web/v1/readme.md <<'EOF'

> **Status (2026-10):** no longer deployed. multibodysystemsdynamics.com is served from `web-sleek/`.
> Glossary, case studies, FAQ and legal content were ported there and are edited in `web-sleek/src/content/`.
> `decisions/`, `programmatic-formbricks/`, `playbooks/` and `requirements/` remain the reference here.
EOF

git commit -am "Mark web/v1 as not deployed; content lives in web-sleek" && git push

git -C external/mbsd-core fetch --tags && git -C external/mbsd-core checkout v0.8.0
git rm app/vendor/mbsd-0.7.0-py3-none-any.whl
(cd external/mbsd-core && uv build --wheel --out-dir ../../app/vendor)
# then bump app/mbsd.pin.json (version + ref) and the MBSD_WHEEL line in app/Makefile
git add external/mbsd-core app/vendor app/mbsd.pin.json app/Makefile && git commit -m "Bump mbsd to
0.8.0"

Energy

Since I read a post about aerothermics, Ive been gathering DHT22 data for few experiments: insulation, modelling sun rays going to a particular home considering the surroundings like mountains or neighbouring buildings

https://github.com/JAlcocerT/poc/blob/main/iot-dashboard-v2/z-setups/1s/components-1s.json

1. Model the balcony recess in Blender — my pick. The scene currently has no loggia at all: 21E is an extruded box with a red marker where the balcony should be. But the recess now drives 45% of the answer, and the 40.9° profile cap is doing that work invisibly. Cutting the real 1.5 m × 2.5 m opening into the facade would let you look at a render of the sunlit patch on your own floor and say “yes, that’s where it reaches” or “no, it comes further in”. That’s the verification the number can’t give itself, and it’s the same job the 21F A/B render did.

2. The seasonal animation. Twelve frames at 17:00 through the year, showing 21F’s shadow climbing your facade and the roof shadow retreating. The counterintuitive headline — March and September are your worst months, December is among your best — is hard to believe in a table and obvious in a loop. create_ffmpeg_concat.py is already there.

3. Nothing to do but keep logging, for two things. The τ fit needs winter. And the drift test needs September–October, when the departure moves 68 min instead of 13.

4. A window-open detector from the humidity channels — you’re already logging humidity on both sensors, and an opening shows as a step in both. That would let the winter fit exclude ventilated periods rather than being poisoned by them. Worth building, but it’s an enabler with nothing to test against until winter, so I’d hold it.

Want me to start on the recess model? It’s the one that makes your observations more useful rather than just prettier — if the render disagrees with what you see on the floor, that tells us the soffit or depth is off before you’ve spent weeks logging against a wrong cap.

I got to know via T3 desktop that the flat next to me, 21F, costs me a consistent 2 h+ every clear day

https://github.com/JAlcocerT/poc/tree/main/building-geo-pl

cd ~/Desktop/poc/building-geo-pl
make pull-data

My TP4056 setup with the ESP32 x DHT11 suffer recently from a full cloudy week.

I measured the 18650 voltage and it was 3.5V

I was wondering how much solar is enough for a micro-controller. Now I know.

Surprise, energy and geolocation matters :O

sqlite3 -header -column /home/jalcocert/poc/iot-rpi-dht-insulation/ingester/data/readings.sqlite "SELECT device, metric, value, topic, received_at, received_ms FROM readings WHERE device='esp32' ORDER BY received_ms DESC LIMIT 1;"

sqlite3 -header -column /home/jalcocert/poc/iot-rpi-dht-insulation/ingester/data/readings.sqlite "WITH intervals AS (SELECT device, topic, received_ms - LAG(received_ms) OVER (PARTITION BY topic ORDER BY received_ms) AS delta_ms FROM readings WHERE device IN
  │ ('esp32','pico')), normal AS (SELECT * FROM intervals WHERE delta_ms BETWEEN 1 AND 600000), counts AS (SELECT device, topic, ROUND(delta_ms/1000.0) AS seconds, COUNT(*) AS occurrences, ROW_NUMBER() OVER (PARTITION BY device, topic ORDER BY COUNT(*) DESC,
  │ ROUND(delta_ms/1000.0)) AS rn FROM normal GROUP BY device, topic, ROUND(delta_ms/1000.0)) SELECT n.device, n.topic, COUNT(*) AS intervals, ROUND(AVG(n.delta_ms)/1000.0,2) AS avg_seconds, ROUND(MIN(n.delta_ms)/1000.0,2) AS min_seconds, ROUND(MAX(n.delta_ms)/1000.0,2) AS..............

• ESP32: approximately every 64 seconds (average ~67 seconds).
• Pico W: approximately every 60 seconds.

Each device sends temperature and humidity as separate MQTT messages during each cycle. Long offline gaps were excluded.

Based on the observed ~64-second cycle, I’d infer:

• ~60 seconds deep sleep
• ~4 seconds booting, reconnecting to Wi‑Fi/MQTT, reading and publishing
• ~1,350 cycles/day
• ~1.5 hours/day awake

Assuming 80–120 mA average while awake and near-ideal deep sleep:

Daily charge ≈ 120–180 mAh Daily energy ≈ 0.40–0.60 Wh

A normal ESP32 development board’s regulator, USB chip and LEDs may raise this to roughly:

≈ 0.5–0.9 Wh/day ≈ 140–260 mAh/day from a 3.7 V battery

So my practical estimate is around 0.6 Wh/day. A 2,000 mAh Li-ion battery would likely last approximately 7–12 days after conversion losses.

The ESP32 chip itself draws about 10 µA in deep sleep, but Wi‑Fi receive uses ~95–100 mA and transmission peaks at 180–240 mA.

Espressif ESP32 datasheet

The frequent Wi‑Fi reconnections dominate consumption.

Extending sleep from 1 minute to 5 minutes could reduce daily usage by roughly 75–80%.

The Pico W appears to remain connected to Wi‑Fi between its one-minute publications. For that setup, I’d estimate:

• Wi‑Fi power saving enabled: ~20–35 mA average
• Power saving disabled/busy loop: ~40–70 mA average
• Likely daily energy: ~2.5–6 Wh/day
• Practical midpoint: ~4 Wh/day

That is roughly 5–8× your deep-sleeping ESP32.

The Pico W’s CYW43439 radio can average below 1.3 mA in Wi‑Fi power-save mode, but active receive uses ~37–43 mA and transmission can peak above 270 mA; the RP2040 and board add their own consumption.

A 2,000 mAh Li-ion might therefore last only around 1–3 days.

If the Pico disconnects and enters genuine low-power sleep between readings, consumption could be reduced substantially.

If the Pico W truly deep-sleeps for 60 seconds, powers down the Wi‑Fi chip, then wakes and reconnects, I’d estimate:

Awake/reconnecting: 4–6 seconds per cycle Daily consumption: ~0.4–0.9 Wh Battery usage: ~120–240 mAh/day at 3.7 V

A practical midpoint is ~0.6 Wh/day, similar to your ESP32. A 2,000 mAh battery might last roughly 8–14 days.

Important: RP2040 deep sleep is around 180 µA, but the CYW43439 radio must also be explicitly powered down; otherwise consumption will be much higher.

Raspberry Pi documentation

Increasing the sleep interval would make a large difference:

• Every 1 minute: ~0.6 Wh/day
• Every 5 minutes: ~0.15–0.25 Wh/day
• Every 15 minutes: ~0.07–0.15 Wh/day

Wi‑Fi reconnection, rather than the sensor reading or MQTT publication, dominates the energy usage.

After having these DHT for several weeks inside and outside home, now i can do per hour checks of T and H:

  sqlite3 -header -column \
  /home/jalcocert/poc/iot-rpi-dht-insulation/ingester/data/readings.sqlite \
  "WITH hourly AS (
    SELECT
      device,
      metric,
      strftime('%Y-%m-%d %H:00:00', received_at) AS hour_bucket,
      AVG(value) AS avg_value
    FROM readings
    WHERE received_ms >= (strftime('%s','now') - 7*24*60*60)*1000
      AND topic IN (
        'esp32/temperature/dht11',
        'esp32/humidity/dht11',
        'pico/temperature/dht22',
        'pico/humidity/dht22'
      )
    GROUP BY device, metric, hour_bucket
  ),
  paired AS (
    SELECT
      e.hour_bucket,
      e.metric,
      e.avg_value AS esp32_value,
      p.avg_value AS pico_value
    FROM hourly e
    JOIN hourly p
      ON p.hour_bucket = e.hour_bucket
     AND p.metric = e.metric
    WHERE e.device = 'esp32'
      AND p.device = 'pico'
  )
  SELECT
    hour_bucket,
    ROUND(MAX(CASE WHEN metric='temperature'
      THEN esp32_value END), 2) AS esp32_temp,
    ROUND(MAX(CASE WHEN metric='temperature'
      THEN pico_value END), 2) AS pico_temp,
    ROUND(MAX(CASE WHEN metric='temperature'
      THEN esp32_value-pico_value END), 2) AS temp_diff,
    ROUND(MAX(CASE WHEN metric='humidity'
      THEN esp32_value END), 2) AS esp32_humidity,
    ROUND(MAX(CASE WHEN metric='humidity'
      THEN pico_value END), 2) AS pico_humidity,
    ROUND(MAX(CASE WHEN metric='humidity'
      THEN esp32_value-pico_value END), 2) AS humidity_diff
  FROM paired
  GROUP BY hour_bucket
  HAVING COUNT(DISTINCT metric) = 2
  ORDER BY hour_bucket;"

  sqlite3 -header -column \
  /home/jalcocert/poc/iot-rpi-dht-insulation/ingester/data/readings.sqlite \
  "WITH RECURSIVE hours(hour_bucket) AS (
    SELECT datetime(
      strftime('%Y-%m-%d %H:00:00','now'),
      '-167 hours'
    )
    UNION ALL

    SELECT datetime(hour_bucket, '+1 hour')
    FROM hours
    WHERE hour_bucket < strftime('%Y-%m-%d %H:00:00','now')
  ),
  counts AS (
    SELECT
      strftime('%Y-%m-%d %H:00:00', received_at) AS hour_bucket,
      device,
      COUNT(*) AS row_count
    FROM readings
    WHERE received_ms >=
          (strftime('%s','now') - 7*24*60*60)*1000
      AND device IN ('esp32','pico')
    GROUP BY hour_bucket, device
  )
  SELECT
    h.hour_bucket,
    CASE WHEN COALESCE(MAX(
      CASE WHEN c.device='esp32' THEN c.row_count END
    ),0) > 0 THEN 'yes' ELSE 'no' END AS esp32_pushed,

    COALESCE(MAX(
      CASE WHEN c.device='esp32' THEN c.row_count END
    ),0) AS esp32_rows,

    CASE WHEN COALESCE(MAX(
      CASE WHEN c.device='pico' THEN c.row_count END
    ),0) > 0 THEN 'yes' ELSE 'no' END AS pico_pushed,

    COALESCE(MAX(
      CASE WHEN c.device='pico' THEN c.row_count END
    ),0) AS pico_rows

  FROM hours h
  LEFT JOIN counts c ON c.hour_bucket = h.hour_bucket
  GROUP BY h.hour_bucket
  ORDER BY h.hour_bucket;"

Yep, im keeping that in the original 60s picow script as a feature, which I detect with the led always ON, not as a bug to know when my ISP is tricking me ;)

But I tweaked the esp32 logic yet again, but keeping this robust deep sleep and wifi reconnections as that one is outside home and I would not realize as quick that Id need to unplug and plug after a router connection issue

To push the new script and learnings:

cd iot-rpi-dht
make deepersleep-upload PORT=/dev/ttyACM0 #10 min interval now

Use mosquitto_sub to watch MQTT messages live:

#mosquitto_sub -h 127.0.0.1 -p 1883 -t '#' -v
#Only monitor the ESP32:
mosquitto_sub -h 127.0.0.1 -p 1883 -t 'esp32/#' -v
#
#docker run --rm --network host eclipse-mosquitto:2 mosquitto_sub -h 192.168.1.2 -p 1883 -t 'esp32/#' -v

Or both devices:

mosquitto_sub -h 127.0.0.1 -p 1883 \
  -t 'esp32/#' \
  -t 'pico/#' \
  -v

This allow the esp32 to push sensor info when properly connected to your wifi:

make deepersleep-upload PORT=/dev/ttyACM0

Crops - Agrotech

The BoM I put together and simulated back in April worked!

After getting the watering setup PoC working, I wanted to tinker with the esp32 wifi connection: beyond the wifimanager

The goal, get all integrated in this user-friendly DIY custom dashboard:

cd ./poc/iot-dashboard-v2
#sudo docker stop qbittorrent

https://github.com/JAlcocerT/poc/blob/main/iot-dashboard-v2/z-learnings-migration.md

Instead of the regular Zigbee config that includes a mqtt server I needed this one to work with my EMQX like so

In the v2 dashboard, use the new “Pump control & schedule” panel to:

• Run a confirmed 0.5–5 second pulse
• Stop the pump
• Refresh its status
• Create/cancel persistent one-shot schedules
• Review recent commands and outcomes

See these CLI equivalents in the makefile to my initial verions: /home/jalcocert/Desktop/poc/iot-esp-water/esp32-wifi and /home/jalcocert/Desktop/poc/iot-esp-water/esp32-bms-prepwork/esp32-bms-mosfet

make pump-status
make pump-pulse PULSE_MS=3000
make pump-off

# docker run --rm --network host eclipse-mosquitto:2 \
#   mosquitto_pub -h 192.168.1.2 \
#   -t esp32/pump/cmd \
#   -m 'pulse:3000'
#Or from esp32-wifi:
# make pub-off

make pump-schedule RUN_AT="2026-09-28 08:00" PULSE_MS=3000
make pump-cancel SCHEDULE_ID=1
make pump-schedules

http://192.168.1.2:3038/?range=90d

/C:/Users/j–e-/Desktop/poc/iot-esp-water/esp32-wifi/wifi_manager_pump/wifi_manager_pump.ino:48

C:/Users/j–e-/Desktop/poc/physics-electronics/voltage-divider/scripts-microcontrollers/firmware-esp32 esp32-dht11-mqtt-emqx-battery-monitor.cpp

C:\Users\j–e-\Desktop\poc\iot-esp32-mppt\ina226-CN3722

mosquitto_sub -h 192.168.1.2 -p 1883 -t ’esp32/mppt/#’ -v mosquitto_sub -h 192.168.1.2 -p 1883 -t ’esp32/#’ -v

• DHT11 DATA → D4 / GPIO4
• Voltage-divider midpoint → D34 / GPIO34
• Pump driver → GPIO23
• INA226 SDA → GPIO21
• INA226 SCL → GPIO22
• INA226 VCC → 3V3
• INA226 GND → GND
• Shared panel/divider ground
• INA226 address 0x40 and 0.1 Ω shunt

esp32/mppt/status partial-measurement esp32/mppt/panel/voltage 10.278 esp32/mppt/charge/voltage 0.000 esp32/mppt/charge/current -0.005 esp32/mppt/charge/power -0.000 esp32/mppt/status partial-measurement

make upload PORT=COM7 WIFI_SSID=“Piszymsiu” WIFI_PASSWORD=“whateveryourwifipwdis123”

sqlite3 -header -csv -readonly /home/jalcocert/poc/iot-dashboard-v2/data/readings.sqlite \
"SELECT datetime(received_ms / 1000, 'unixepoch', 'localtime') AS time_local,
    value AS panel_voltage_v
  FROM readings
  WHERE topic = 'esp32/panel/voltage'
    AND received_ms >= unixepoch('now', 'localtime', 'start of day', 'utc') * 1000
    AND received_ms <  unixepoch('now', 'localtime', 'start of day', '+1 day', 'utc') * 1000
  ORDER BY received_ms ASC;" > "panel_voltage_$(date +%F).csv"

scp jalcocert@192.168.1.2:/home/jalcocert/panel_voltage_2026-10-09.csv .

Sun Rays Modelling

To take the sun ray modelling further, I used the Pi4 with its camera:

rpicam-still --nopreview --timeout 2000 --output camera-tests/snapshot-20260930-151243.jpg
#scp jalcocert@192.168.1.18:/home/jalcocert/camera-tests/snapshot-20260930-151243.jpg .
#rsync -avz jalcocert@192.168.1.18:/home/jalcocert/camera-tests/latest.jpg .
make latest
#scp jalcocert@192.168.1.18:/home/jalcocert/camera-tests/captures/snapshot-20261002-084222.jpg .

I recommend:

• Now / experiments: shell loop inside tmux
• Later / always-on: systemd timer or cron

I added the tmux-friendly version here:

camera-tests/capture-every-5min.sh camera-tests/Makefile

Use it like this on the Pi:

cd /home/jalcocert/camera-tests make tmux-start

It will capture one rotated image every 5 minutes into:

/home/jalcocert/camera-tests/captures/

and refresh:

/home/jalcocert/camera-tests/latest.jpg

Useful commands:

make tmux-attach make tmux-stop make list

Change interval, for example every 60 seconds:

make tmux-start INTERVAL_SECONDS=60

From Windows, you can keep copying the newest image with:

scp jalcocert@192.168.1.18:/home/jalcocert/camera-tests/latest.jpg .

─ Worked for 5m 47s ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────

› so that will run until i stop it?

• Yes.

make tmux-start starts a detached tmux session running the capture loop. It will keep taking a snapshot every 5 minutes until one of these happens:

• you stop it with:

cd /home/jalcocert/camera-tests make tmux-stop

• the Pi reboots or powers off
• the script crashes because of a camera/system error
• the tmux session is killed manually

You can watch it live with:

make tmux-attach

Then detach without stopping it by pressing:

Ctrl+b d

To confirm it is running:

tmux ls

or:

tail -f /home/jalcocert/camera-tests/capture-loop.log

I added disk safeguards to the capture loop.

Current Pi disk status:

28G total, 24G used, 3.4G free, 88% used

The loop now defaults to:

MAX_CAPTURES=576 MIN_FREE_MB=1024 LOG_MAX_LINES=2000

At one image every 5 minutes, 576 captures is about 2 days of history. After each successful capture, the script:

1. deletes oldest captures if there are more than MAX_CAPTURES
2. deletes oldest captures if free disk space drops below MIN_FREE_MB
3. trims the log so it does not grow forever

Use:

cd /home/jalcocert/camera-tests make tmux-start

More conservative, keep only 1 day and at least 2GB free:

make tmux-start MAX_CAPTURES=288 MIN_FREE_MB=2048

Check space anytime:

make disk

If a capture loop is already running, restart it to pick up the new safeguards:

make tmux-stop make tmux-start

While experimenting wanted to also understand sun and earth positioning: https://solar-trajectory.pages.dev/

Trying to avoid homography

FPV Telemetry

• https://jalcocert.github.io/JAlcocerT/dji-tello-python-sdk/#tello-x-flutter
• https://jalcocert.github.io/JAlcocerT/desktop-apps-with-rust/

Ive also shipped a Kotlin version that you can get via obtanium.

fpv dron prop vortex https://www.youtube.com/shorts/hgqM9Z0d6QU

https://www.youtube.com/@rctestflight https://www.youtube.com/@fpv-geek https://www.youtube.com/@JoshuaBardwell/videos https://www.youtube.com/@opendrone

Got to know about https://opendrone.be/ so I applied my ./poc/fpv-kpi to see if the build make sense specially as all repos are opened https://github.com/orgs/OpenDrone-hw/repositories

MPU acelerometer

There are many 3-axis accelerometers that you can use with the Raspberry Pi Pico.

Some of the most popular options include:

MPU-6050: This is a popular and versatile accelerometer that is also compatible with the Raspberry Pi Pico.

It has a wide range of features, including a built-in gyroscope.

biblioman09

Others

Attract Convert Deliver

There are changes going on in all orgs.

Some call them: design-sell-deliver-enable

But they come down to the same.

1. Have been doing changes and additions to the ebooks

Anytime you want https://ebooks.jalcocertech.com/ * There is Free DIY: IoT and electronics!

The idea here is: if the quality of the free is so good, how will it be the quality of the paid consulting or DFY services?

2. Have added the latest tech-talks to https://consulting.jalcocertech.com/presentations/techtalk-from-iot-to-big-data-engineering/ppt

Leeeeads

Oh yea, the leads!

What am i doing about that?

It’s all about having a proper leads pipeline:

cd ./fossengineer/
#https://github.com/JAlcocerT/poc/tree/main/genbi-energy-solutions/waitlist
cd ./poc/tree/main/genbi-energy-solutions/waitlist
#https://github.com/JAlcocerT/jalcocertech-core/tree/main/leads-hub/hub
#cd ./jalcocertech-core/
sqlite3 -readonly -json data/leads.db "SELECT received_at, site, kind, email, json(raw) AS raw FROM leads ORDER BY received_at DESC LIMIT 10;"

After you get them, you enrich them as I applied to myself here

codex --search

I got to know about cloudflare workers KV while improving Independent engineering for systems in motion.

Once i configured the CNAMES:

dig multibodysystemsdynamics.com any
#ping app.multibodysystemsdynamics.com #app-mbsd.pages.dev

Some time ago i made https://multibodysystemdynamics.pages.dev/

http://app.multibodysystemsdynamics.com/ -> app-mbsd.pages.dev

From ~/Desktop/mbsd-framework/multi-body/web-sleek. First, point the config at the v1 project, so the KV binding and make deploy target it:

  sed -i 's/^name = "multibodysystemsdynamics-site"/name = "multibodysystemsdynamics"/' wrangler.toml
  sed -i 's/^CF_PAGES_PROJECT ?= multibodysystemsdynamics-site/CF_PAGES_PROJECT ?=
  multibodysystemsdynamics/' Makefile
  sed -i 's/^CF_PAGES_BRANCH ?= main/CF_PAGES_BRANCH ?= master/' Makefile
  grep -E "^name|^CF_PAGES" wrangler.toml Makefile          # check all three changed

    Deploy (legal check, build, upload to production):

  make deploy

  Check the live domain:

  for p in / /tools/ /glossary/grashof-condition/ /case-studies/practica/ /legal/privacy/ /rss.xml; do
    curl -s -o /dev/null -w "%{http_code} $p\n" "https://multibodysystemsdynamics.com$p"; done
  curl -s https://multibodysystemsdynamics.com/ | grep -o "<title>[^<]*"   # should show the new title

  Commit the config change:

  git -C .. add web-sleek/wrangler.toml web-sleek/Makefile
  git -C .. commit -m "web-sleek: deploy to the multibodysystemsdynamics project (apex domain)" && git
  -C .. push

Rollback if needed

  Dashboard → Workers & Pages → multibodysystemsdynamics → Deployments → the v1 deployment (de5c9561,
  branch master) → ⋯ → Rollback to this deployment. The apex serves v1 again within seconds.

Notes

• The KV binding moves along automatically. It’s defined in wrangler.toml, so the deploy applies it to the v1 project, and the contact form keeps writing to the same LEADS namespace.
• The multibodysystemsdynamics-site project becomes redundant. Keep it as a staging copy, or delete it later with npx wrangler pages project delete multibodysystemsdynamics-site.
• v1’s own files stay in git under web/, untouched, as you wanted. Only the served deployment changes.
• www still doesn’t exist; adding a www CNAME plus a redirect rule to the apex remains an optional extra.

I can run these myself and run the live checks if you’d like. Say so and I’ll do it.

cd ./multi-body/web-sleek #new https://multibodysystemsdynamics.pages.dev #old kineo https://multibodysystemdynamics.pages.dev/
make deploy #make dev
#curl -s https://multibodysystemsdynamics.com/rss.xml | head -10
#ping www.multibodysystemsdynamics.com
make leads-show

https://multibodysystemsdynamics.com/rss.xml

whats KV?

KV is Cloudflare Workers KV, a simple key-value store that lives on Cloudflare’s network, roughly a giant dictionary where each entry is a key (a name) plus a value (any text, here JSON).

How the site uses it: when someone sends the contact form or requests the ebook, the site’s small backend function (web-sleek/functions/api/contact.ts) saves the submission as one entry:

key:   enquiry:2026-10-04T09:30:12.000Z:3f2a…      (or ebook:… for ebook requests)
value: {"name":"Ada","email":"ada@acme.com","project":"Solver validation",
        "details":"…","receivedAt":"…","country":"ES","referer":"…"}

LEADS is the namespace (think of it as one dictionary) I created for these. It’s bound to the site in wrangler.toml, which is how the function finds it.

Reading your leads:

• Dashboard: Storage & Databases → KV → LEADS shows every entry.
• Command line:

npx wrangler kv key list --namespace-id e4e917c99f9d46ecab4c2bf25074090f --remote
npx wrangler kv key get "<key from the list>" --namespace-id e4e917c99f9d46ecab4c2bf25074090f
--remote

Why KV for this:

• It’s free at your scale: the free tier allows 1,000 writes and 100,000 reads a day, plus 1 GB of storage. That’s far more than a contact form will ever use.
• No server or database to run. It’s built into the same Cloudflare account and deployment as the site.
• Nothing is lost. Even if you never set up notifications, every enquiry is kept.

Limitations:

• It doesn’t notify you. Leads just sit there until you look, which is why I suggested adding a webhook (Slack, Discord, n8n) for a ping per enquiry. The function supports both at once: stored in KV and sent to the webhook.
• It’s a store, not a CRM. No search, statuses or replies. At higher volume you’d forward leads to a CRM or an email inbox instead.
• It’s eventually consistent: a new entry can take up to a minute to appear everywhere. That doesn’t matter for a contact form.

The privacy policy already mentions “form storage (Workers KV)” for exactly this.

I’ve already validated it twice: a test submission through the live domain returned 202 and appeared in LEADS, and I deleted it afterwards. Here’s how to check it yourself, end to end.

1. Send a real submission through the site. Open https://multibodysystemsdynamics.com/contact/, fill in the form with your own email, and press Send enquiry. You should see “Thanks — your enquiry has been received.” If the backend failed, you’d instead see “The form could not be sent right now. Send it by email instead.” For the ebook path, do the same at /education/.

2. Look at it in the dashboard. Go to: Storage & Databases → KV → LEADS → KV Pairs

Your entry appears as enquiry:2026-10-04T… (or ebook:…). Click it to see the JSON. It can take up to a minute to show up.

3. Or from the terminal (in multi-body/web-sleek): NS=e4e917c99f9d46ecab4c2bf25074090f

# list all leads (key + email/project metadata)
npx wrangler kv key list --namespace-id $NS --remote

# read one in full
npx wrangler kv key get "enquiry:2026-10-04T…" --namespace-id $NS --remote

# delete your test entry afterwards
npx wrangler kv key delete "enquiry:2026-10-04T…" --namespace-id $NS --remote

#npx wrangler kv key get "enquiry:2026-10-04T10:12:41.585Z:8612a6e5-07de-4ec5-831a-33e2c90decfb" --namespace-id $NS --remote
#{"kind":"enquiry","name":"cerdo","email":"jesalctag@gmail.com","project":"Mechanism modeling","details":"eres un cerd

4. Quick backend-only check (no browser), to tell a page problem apart from a storage problem:

curl -s -w " %{http_code}\n" -X POST https://multibodysystemsdynamics.com/api/contact \
  -H "Content-Type: application/json" \
  -d '{"name":"KV check","email":"you@example.com","project":"Something else","details":"test"}'
# expect: {"ok":true} 202

Reading the responses:

┌─────────────────────┬─────────────────────────────────────────────────────────────────────────┐ │ Response │ Meaning │ ├─────────────────────┼─────────────────────────────────────────────────────────────────────────┤ │ 202 {“ok”:true} │ Saved to KV │ ├─────────────────────┼─────────────────────────────────────────────────────────────────────────┤ │ 422 invalid_fields │ Validation rejected it (bad email or missing fields): the backend works │ ├─────────────────────┼─────────────────────────────────────────────────────────────────────────┤ │ 503 not_configured │ The KV binding isn’t attached to the deployment │ ├─────────────────────┼─────────────────────────────────────────────────────────────────────────┤ │ 502 delivery_failed │ KV (and the webhook, if set) both failed to write │ └─────────────────────┴─────────────────────────────────────────────────────────────────────────┘

What doesn’t happen yet: nothing emails you when a lead arrives. Leads wait in KV until you look.

Since your mail runs through Proton, the simplest notification is a webhook, for example a Discord or Slack channel.

Give me the URL and I’ll set it as LEAD_WEBHOOK_URL. Each enquiry is then saved to KV and posted to the channel.

To read every lead in one go, without copying keys around:

for k in $(npx wrangler kv key list --namespace-id $NS --remote | python3 -c "import
json,sys;[print(x['name']) for x in json.load(sys.stdin)]"); do
  echo "== $k"; npx wrangler kv key get "$k" --namespace-id $NS --remote; echo; done

I’ve added make targets for the leads, tested each one against the live KV, and committed them (d5aff35). Run them from multi-body/web-sleek:

make leads # table: time, kind, email, project make leads-show # every lead in full (JSON) make leads-keys # raw keys, for copy-paste make lead KEY=enquiry:… # one lead make lead-delete KEY=… # delete one make leads-test # send a test enquiry to the live site (expect 202) make leads-clean-tests # delete test leads (emails ending in @example.com)

Example output:

received (UTC) kind email project 2026-10-04 10:12:41 enquiry abcd@gmail.com Mechanism modeling

1 lead(s). Show one: make lead KEY= keys: make leads-keys

• No more install prompt: wrangler is now a dev dependency of web-sleek.
• No ID to remember: the scripts read the namespace ID from wrangler.toml.
• Test cleanup: make leads-test sends from kv-check@example.com, so make leads-clean-tests removes it along with any other @example.com test. I already ran that, which also removed your earlier you@example.com curl test. Only your Gmail entry remains; delete it with make lead-delete KEY=$(make -s leads-keys | head -1) when you’re done with it.
• The commands are in the README under “Leads (contact + ebook forms)”.

Beyond the PWA, the wheel and KV, here’s what we did and what came out of it.

Release and OSS:

• Reviewed mbsd-core and mbsd-examples before 0.7.0. Caught the stale v0.6.0 checkout lines in both READMEs, and the release-order trap: tag core before pushing examples, or the examples CI breaks.

• Found a real core bug: grashof_class() labels a Grashof linkage with the shortest output link as “double-rocker”. It’s still open, worth an issue and a test for 0.8.

• Checked the new spatial kinematics maths and found it correct.

  The website

  • Turned an invisible page into an indexable site. The old landing page was all JavaScript-injected, with nothing for crawlers. It’s now static Astro with 25 pages, sitemap, canonical tags, OG images and structured data.
  • Built “instant preview, full solver one click away”. The tool pages run closed-form maths in about 11 KB of JS, then hand off to the real solver with the parameters in the URL. Both sides give the same numbers (49.6°, μ 46.1–99.2°).
  • Made the code samples trustworthy. The site’s code sample called a method that doesn’t exist; now every Python sample on the site runs against your core (make snippets), and the hero’s residual is computed instead of hardcoded.
  • Migrated v1 without losing anything. All 17 indexed URLs kept, broken images replaced with real figures, and a factual error fixed (the swapped Grashof table). Your old web/ is preserved as you asked.

  The app

  • Found it was silently broken. import mbsd failed in Pyodide because raw source has no package metadata. The wheel fixed that and pinned what visitors run to a tagged release.
  • Made “offline” honest. The service worker now caches everything atomically, prompts for updates, and uses versioned immutable paths. Tested with an actual offline reload, not assumed.
  • Kept the demo models in one place. They became one real Python file, tested in CPython and in Pyodide, instead of a string inside TypeScript.

  Infrastructure and process

  • Put web and app in one private repo, with the OSS pulled in from outside. Submodules at release tags plus a 54 KB committed wheel: no PyPI needed yet, and builds never touch the private workbench.
  • Shipped to production with no downtime. Deploying into the existing project instead of moving the domain kept rollback to a single click.
  • Replaced the Formbricks flow with a tiny Pages Function. Store in KV, optionally ping a webhook, fall back to a pre-filled email, with a honeypot against bots. Less third-party code, and you own the data.
  • Debugged Cloudflare’s quirks along the way: production branch names, the www 522 (a hostname not registered on the project, not a DNS issue), email obfuscation hiding addresses from grep, and the edge rewriting sw.js cache headers.

  Legal and privacy

  • Replaced two deployed “Do not deploy this stub” pages with an accurate privacy policy and terms, plus a build guard so placeholders can’t reach production again.
  • Linked the site to JAlcocerTech without exposing your home. The insight: a visible NIP leads to the CEIDG address. GDPR is covered now; the virtual-office step is written down for later.

  Lessons worth keeping

  1. Verify, don’t assume. Several of the biggest finds (the broken app, broken images, deployed stubs, the non-running sample) came from actually running or fetching things rather than reading code.
  2. Same numbers everywhere. Core, examples, site preview and app all agree on the same test cases, and that agreement is your brand.
  3. Guards over reminders. The version check, check-legal and the snippet runner each turn a “remember to…” into a build that refuses to proceed.
  4. Separate what’s public from what’s private on purpose. OSS engine, private product, tagged releases crossing the boundary, and no personal data leaking through a side channel like the NIP.

  Still open, in rough order:

  • the www → apex redirect rule;
  • a lead webhook, so new enquiries notify you;
  • Web Analytics;
  • GitHub Actions deploys;
  • the virtual office address and NIP;
  • the email-gated report in the app, which is the piece that turns tool usage into contacts.

Webs

I sunseted all custom diy websites from 2024, as they churned anyways

Created this thought: https://github.com/JAlcocerT/poc/tree/main/pwa-margincms

As a PWA that you can use offline via chrome: https://margin-cms.pages.dev/

With simple gg syncronization via PAT.

Oh, and you also have the free web audits to show that yours have a problem: https://webaudit.jalcocertech.com/

Ive started unifying my workflows across brands.

Initially was doing a custom mail tests with fastapi x sqlite x mailpit here

Site: multibodysystemsdynamics.com Hosting: Cloudflare Pages (direct upload) Stack: Astro + Pages Function Can it use the KV namespace directly?: ✅ already does ──────────────────────────────────────── Site: www.jalcocertech.com Hosting: Cloudflare Pages, git-connected to JAlcocerT/jalcocertech Stack: Astro Can it use the KV namespace directly?: ✅ possible (same account) ──────────────────────────────────────── Site: fossengineer.com Hosting: Cloudflare Pages, git-connected to fossengineer1/fossengineerpapermod Stack: Hugo (PaperMod) Can it use the KV namespace directly?: ✅ possible, but Hugo has no server code of its own

What I’ve confirmed so far (from the repos, plus the parts of the earlier live check that were reliable):

Property: www.jalcocertech.com (jalcocertech) Stack / hosting: Astro, Cloudflare Pages Lead capture today: Footer form posts to formsubmit.co, emailing contact@jalcocertech.com Notes: The contact page form has no action and no script, so submitting it reloads the page with name

and email in the URL and nothing is sent. No privacy page in the repo ──────────────────────────────────────── Property: consulting.jalcocertech.com (selfhosted-landing) Stack / hosting: Astro in Docker on the homelab via a Cloudflare tunnel Lead capture today: A Formbricks survey link plus a cal.com booking embed; also a chatbot widget Notes: Has privacy and terms pages. My first search matched “gtag” (Google Analytics); not yet confirmed whether it really loads ──────────────────────────────────────── Property: ebooks.jalcocertech.com (1ton-ebooks) Stack / hosting: Astro, deployed to Pages Lead capture today: No forms, just cal.com links Notes: Its site setting is ebooks.jalcocert.com (not jalcocertech), so the canonical URLs and sitemap

may point at the wrong domain. Was confirming. No legal pages ──────────────────────────────────────── Property: webaudit(s).jalcocertech.com (poc_webs_magnet) Stack / hosting: Express + PocketBase, Turnstile, Docker Lead capture today: Audit form saves name, email and website into PocketBase (leads, audits) and a CSV file Notes: The checkbox combines “results are sent by email” with “I accept marketing comms”, which isn’t

valid consent: marketing consent must be separate and optional. leads.csv and manual-leads/*.md are committed to git (the repo is private, so it isn’t exposed, but personal data is in git history)

All four repos are private.

This matters for leads-hub because each property captures leads differently today: formsubmit, Formbricks, PocketBase, and one broken form. That’s an argument for the central Worker (option C). It would also fix the www contact form and the bundled consent along the way.

There’s also an unrelated issue to fix: the apex jalcocertech.com resolves publicly to 192.168.1.11, a private home-network address.

Visitors typing jalcocertech.com without www can’t reach anything, and it reveals an internal IP.

It’s probably an A record meant for your homelab. Point the apex at the site, and/or add the same www-style redirect we just did for MBSD.

The key design change: one central forms Worker instead of a function per site

Copying the Pages Function into every site would mean four copies drifting apart, and it can’t work for leadarchitect.org (Firebase) or easily for Hugo.

Instead:

any site (Astro, Hugo, Firebase…) ──fetch──► https://forms.jalcocertech.com/v1/submit (one Cloudflare Worker) ├─ checks Origin against an allowlist of your sites ├─ validates per form kind (enquiry / ebook / newsletter) ├─ honeypot (+ Turnstile later if spam appears) ├─ KV write: :::, 24-month TTL, consent record └─ ntfy alert (site · kind · email domain)

homelab leads-hub (pull-only, every 10 min) ◄── Cloudflare API, read-only KV token SQLite: leads_raw → enrichment → digest / export / UI

• One codebase and one deploy, with consistent validation, consent handling and alerts on every site.
• Each site only needs a form and ~20 lines of JS. It works for Astro, Hugo and Firebase alike.
• MBSD can keep its current function at first and switch its form to the Worker later, at which point the old function goes.

GDPR across sites (no billing, so the address gap stays acceptable)

1. One privacy-policy template, adapted per site: controller (JAlcocerTech), Cloudflare as processor, the homelab, purposes per form kind, 24-month retention, rights, and UODO for complaints.
2. Newsletter forms need explicit opt-in consent (an unticked checkbox), proof of that consent stored with each lead (policy version and time), and a way to unsubscribe. Enquiry and ebook forms don’t need a consent checkbox; replying to someone who asked is enough of a legal basis.
3. Enrichment and linking the same person across sites must be described in every policy before the pipeline runs: legitimate interest, plus the right to object.
4. Deletion requests: a leads-hub forget command that removes the person from SQLite and KV across all sites.
5. leadarchitect.org on Firebase: check whether it runs Google Analytics. If it sets cookies, that site needs a consent banner.

leads-hub on the homelab

• Separate private repo: a Python pull job, SQLite, and a systemd timer or Docker Compose.
• Tables: leads_raw (unchanged, keyed by the KV key) → people (deduplicated by email across sites) → enrichment (all locally computable: free vs company email domain, MX check, company domain, site and kind history, a simple score).
• Outputs: a weekly ntfy digest, CSV export, and optionally Datasette for a browser interface on the SQLite file. That would sit on the homelab, behind Cloudflare Access.

Proposed phases

1. Forms Worker + key format: build the Worker at forms.jalcocertech.com, write the shared policy template, migrate MBSD’s form, and confirm with a test from each site.
2. leads-hub v0: pull job, SQLite, deletion command, deployment on the homelab.
3. Roll out to jalcocertech.com and fossengineer.com: forms plus updated privacy policies.
4. Enrichment and digest, once the policies say so.
5. leadarchitect.org, after checking its Firebase setup.

A few decisions before I start on phase 1:

1. A central Worker at forms.jalcocertech.com. OK, or would you prefer a different name or domain?
2. Which form kinds per site? For example: enquiry on jalcocertech.com, newsletter on fossengineer.com?
3. Is Datasette fine as the leads interface, or would you rather start with only make commands, like now?
4. The jalcocertech.com apex: should I tell you exactly what to change in DNS, or is the 192.168.1.11 record intentional?

Case Studies

Electronic Design

Yep, i designed and sent for manufacturing recently my first pcb.

There were 3 stages:

1. Electronic Simulation
2. Bread and protoboard testing
3. PCB Design with KiCAD

3D-Design

Governance Consulting

Believe it or not, these are not clear and organizations still have severe governance problems:

• https://ebooks.jalcocertech.com/books/dna/dna-career-skills/#project-management-essentials
• https://ebooks.jalcocertech.com/books/managing-data-projects/faq-project-docs/

HomeLab

This setup is working quite nicely thanks to skills:

Ive made some updates: https://fossengineer.com/contact/ and https://fossengineer.com/privacy/

What’s on you:

• Unsubscribes are manual for now. When someone replies “unsubscribe”, run make leads to find them and make lead-delete KEY=… to remove them. The forget command comes with leads-hub phase 2.

• The quarterly email itself: your subscriber list is make leads (filter fossengineer.com / newsletter). Send from hello@teco.com with recipients in BCC, so subscribers don’t see each other’s addresses.

CF Workers free plan gives: 100,000 requests per day + Up to 10 ms CPU time per request + Community support

Conclusions

Making questions is the first step.

Then its about making good questions, like:

• Stop asking: “Do they see it?
• Start asking: “Is this priced correctly?”

Some orgs are already asking: why do we need a person?

And that makes sense, the info is out there: ppl are realizing that scrum doesnt work, aka is too slow

How long could it last earning 6 figures by prompting an agent?

Now…dark factories will come (even more) to the software/IT sector

AI…employees coming?

Comercial ones like: https://www.lindy.ai/pricing

Why not building a brand before its too late:

Consulting Services

Consulting - Tier of Service

DIY via ebooks

Distilled knowledge via web/ooks with free value.

Im putting together a JAlcocerTech-Core:

flowchart LR
    %% --- Styles ---
    classDef free fill:#E8F5E9,stroke:#2E7D32,stroke-width:2px,color:#1B5E20;
    classDef low fill:#FFF9C4,stroke:#FBC02D,stroke-width:2px,color:#FBC02D;
    classDef mid fill:#FFE0B2,stroke:#F57C00,stroke-width:2px,color:#F57C00;
    classDef high fill:#FFCDD2,stroke:#C62828,stroke-width:2px,color:#C62828;
    classDef bridge fill:#E3F2FD,stroke:#1565C0,stroke-width:3px,color:#0D47A1;

    %% --- Nodes ---
    L0("Free Content
( DIY = $0)"):::free
    L1("Web Audits 🛡️
(Reveals Problem )"):::free
    L11("Tech Blog/Youtube"):::free
    L12("ebooks"):::free
    L13("mbsd framework OSS"):::free
    L14("OSS guides"):::free

    L3("Done With You
(Trade $$ for knowledge)"):::mid
    L4("Done For You
(Trade $$$ for outcomes)"):::high
    L44("GenBI
Shopify PoC"):::bridge
    L45("Real Estate
Funnel Bot"):::bridge
    L46("Energy Solutions
HVAC"):::bridge
    L47("IoT Solutions
Crops"):::bridge
    L48("Weddings
Photo QR"):::bridge

    %% --- Connections ---
    L0 --> L1
    L1 --> L3
    L12 --> L3
    L13 -->|MultiBodySystemsDynamicscom| L3
    L14 -->|FOSS Engineer| L3
    L0 --> L11
    L0 --> L12
    L0 --> L13
    L0 --> L14
    L3 --> L4
    L4 -->|Productized Service| L44
    L4 -->|Productized Service| L45
    L4 -->|Productized Service| L46
    L4 -->|Productized Service| L47
    L4 -->|Productized Service| L48

How can we work together?

One of my favourite converging questions I got this year

The Ways of Working (WoW) of many are far from perfect

And im not even talking about using AI

As the cost and quality of replies gets better, the outcomes depend more on the quality of the questions and our learning rate, proper meta-frameworks adoption

Learn how to delegate the work to agents to anyone

just not your understanding unless you have a team to deploy ideas

Choosing my WoW

The beauty of optionality is that i can choose.

As my career is no longer bottlenecked by permission but by how well I allocate leverage, taking care of my time and boundaries is crucial.

Having a clear game and playbook help to operate this smoothly

When ppl asked me for collaborations, I make sure to cross-check their proposal with a bs detection form i created as a code here and deployed to formbricks

Case Studies

Clarity of Execution

Working in D&A?

Go ask unconfortable questions: smart or it does NOT ship

• https://why-postmortem-checks.pages.dev
• https://pm-pdm-checks.pages.dev

You might not know yet, but you need proper governance.

You cant be an AI first company before you are a data ready team.

To be a data ready team, you need proper RACI model across product, architecture and delivery.

And to even get started: you need to have some kind of logic

Example: a date is not a product definition

If you are still working in a 9-5 while working in your free time to make your business, make sure to have a clear picture of what your game is and a playbook to execute.

FAQ

PIO

In software engineering, operations, and business analysis, framing PIO as Problem, Integration, Outcome creates a sharp model for designing architecture, automating workflows, and writing clear business requirements.

See https://www.seangoedecke.com/tell-agents-the-why/

• Problem: The specific operational bottleneck, system defect, data silo, or manual inefficiency in the current workflow (e.g., “Customer support manually re-keys order data across two legacy databases, causing a 24-hour fulfillment lag”). THE WHY and slightly what

• Integration: The technical connection, automated workflow, API bridge, or architectural change introduced to bridge the gap (e.g., “Deploy an event-driven webhook via an enterprise service bus (ESB) to sync order status in real time”). whats everything/systems that the agent needs? where is the agent going to take info from?

• Outcome: The quantifiable, verifiable metric or end state defining success (e.g., “Order processing time reduced from 24 hours to under 30 seconds; 0% manual data entry errors”). THE WHAT

Shift the conversation away from low-level implementation debates to high-level governance rules, evidence models, and risk ownership.

Strong governance framing that clearly establish:

• Why the control exists.
• What is being assessed.
• What evidence is required.
• What constitutes Pass / Action Required / Unable To Assess.
• Who owns the decision.
• What the assistant can and cannot do.

That separation of responsibilities is usually what directors and architects care about most.

PIO (Problem, Integrations, Outcome)

• Good for: Defining the governance guardrails, evidence boundaries, and goal criteria for AI agents.
• SDLC Role: Replaces or encapsulates heavy BRD/PRD documentation specifically for agentic workflows and automated platforms.
• Key Question Answered: “What specific enterprise problem are we evaluating, what systems hold the truth, and what decision should the agent output?”
• Primary Audience: Directors, DevSecOps leads, enterprise architects, and prompt/agent engineers.
• Core Contents: Problem statement, integrations/evidence sources, evaluation logic (Pass / Action Required), metadata, and authority limits.

the PIO question flow:

• Problem asks the why.
• Integrations ask what evidence and systems prove it.
• Outcome defines what the assessment determines.
• Assessment logic turns evidence into Pass / Action Required / Unable To Assess.
• Agent output packages the result.
• Governance questions feed back into leadership decisions and clarify future versions.

For these Director-style PIOs, Problem → Integrations → Outcome flows more naturally because it mirrors:

• What issue are we solving?
• What data/systems are involved?
• What does the assessment produce?

flowchart LR
    A[Start With Desired State Criterion] --> B[Problem Questions]

    B --> B1[Why does this matter?]
    B --> B2[What risk exists?]
    B --> B3[What cannot be considered aligned?]
    B --> B4[What loopholes or ambiguity must be closed?]

    B1 --> C[Problem Statement]
    B2 --> C
    B3 --> C
    B4 --> C

    C --> D[Integration Questions]
    D --> D1[What systems hold the evidence?]
    D --> D2[What is the primary evidence key?]
    D --> D3[Which standards or policies apply?]
    D --> D4[Which human inputs are needed?]
    D --> D5[Which evidence is preferred when sources conflict?]

    D1 --> E[Evidence / Integration Model]
    D2 --> E
    D3 --> E
    D4 --> E
    D5 --> E

    C --> F[Outcome Questions]
    E --> F

    F --> F1[What are we trying to determine?]
    F --> F2[What evidence proves or disproves conformance?]
    F --> F3[What gaps should be identified?]
    F --> F4[What result should the reviewer receive?]

    F1 --> G[Outcome]
    F2 --> G
    F3 --> G
    F4 --> G

    G --> H[Assessment Logic]
    E --> H

    H --> H1[Pass]
    H --> H2[Action Required]
    H --> H3[Unable To Assess]

    H1 --> I[Agent Output]
    H2 --> I
    H3 --> I
    G --> I

    I --> I1[Reviewer-Ready Artifact]
    I --> I2[Evidence Summary]
    I --> I3[Findings And Risk]
    I --> I4[Remediation Actions]
    I --> I5[Executive Summary]

    I --> J[Governance Questions]
    J --> J1[Who owns the decision?]
    J --> J2[What does the assistant not do?]
    J --> J3[What needs leadership confirmation?]

    J1 --> K[Director Review Position]
    J2 --> K
    J3 --> L[Leadership Decisions Needed]

    L -.feeds back.-> B
    L -.clarifies.-> D
    L -.sets thresholds.-> H

The core questions asked were:

• Problem: Why does this matter? What risk exists? What cannot be considered aligned? What ambiguity must be closed?
• Integrations: What systems hold the evidence? What is the primary evidence key? Which standards or policies apply? Which human inputs are needed?
• Outcome: What are we trying to determine? What evidence proves or disproves conformance? What gaps should be identified?
• Assessment Logic: What is a Pass? What requires Action Required? When is the assessment Unable To Assess?
• Agent Output: What artifact should the reviewer receive? What evidence, findings, risks, remediation actions, and summary should it contain?
• Governance: Who owns the decision? What does the assistant not do? What needs leadership confirmation?

HLD: PIO Question Flow

flowchart LR
    A[Start With Desired State Criterion] --> B[Problem Questions]

    B --> B1[Why does this matter?]
    B --> B2[What risk exists?]
    B --> B3[What cannot be considered aligned?]
    B --> B4[What loopholes or ambiguity must be closed?]

    B1 --> C[Problem Statement]
    B2 --> C
    B3 --> C
    B4 --> C

    C --> D[Integration Questions]
    D --> D1[What systems hold the evidence?]
    D --> D2[What is the primary evidence key?]
    D --> D3[Which standards or policies apply?]
    D --> D4[Which human inputs are needed?]
    D --> D5[Which evidence is preferred when sources conflict?]

    D1 --> E[Evidence / Integration Model]
    D2 --> E
    D3 --> E
    D4 --> E
    D5 --> E

    C --> F[Outcome Questions]
    E --> F

    F --> F1[What are we trying to determine?]
    F --> F2[What evidence proves or disproves conformance?]
    F --> F3[What gaps should be identified?]
    F --> F4[What result should the reviewer receive?]

    F1 --> G[Outcome]
    F2 --> G
    F3 --> G
    F4 --> G

    G --> H[Assessment Logic]
    E --> H

    H --> H1[Pass]
    H --> H2[Action Required]
    H --> H3[Unable To Assess]

    H1 --> I[Agent Output]
    H2 --> I
    H3 --> I
    G --> I

    I --> I1[Reviewer-Ready Artifact]
    I --> I2[Evidence Summary]
    I --> I3[Findings And Risk]
    I --> I4[Remediation Actions]
    I --> I5[Executive Summary]

    I --> J[Governance Questions]
    J --> J1[Who owns the decision?]
    J --> J2[What does the assistant not do?]
    J --> J3[What needs leadership confirmation?]

    J1 --> K[Director Review Position]
    J2 --> K
    J3 --> L[Leadership Decisions Needed]

    L -.feeds back.-> B
    L -.clarifies.-> D
    L -.sets thresholds.-> H

How They Compare in the Lifecycle

Application Across Roles

Why It Works Better Than Standard User Stories

While traditional user stories (“As a [user], I want [feature] so that [benefit]”) focus primarily on end-user features, Problem, Integration, Outcome excels for system-to-system requirements, backend optimizations, and cross-platform workflows because it explicitly forces teams to define how systems talk to each other rather than just describing surface-level behavior.

Example:

• Problem: “As part of our journey toward…” + concrete problems uncovered
• Integrations: systems/standards/portals/dashboards/repos involved
• Outcome: “As a result…” + what the agent will assess/reference/report/
  recommend

For software engineering, operations, and business analysis, several frameworks mirror PIO by structuring problem-solving, architectural choices, and requirements gathering.

Framework Comparison Matrix

Systems & Architecture Frameworks

• C4 Model (Context, Containers, Components, Code)

• Best For: Software architecture and system integration diagramming.

• Focus: Maps complex software architectures at four progressive levels of abstraction, making system connections and boundaries visually explicit.

• Architecture Tradeoff Analysis Method (ATAM)

• Best For: Evaluating software architecture before implementation.

• Focus: Evaluates structural choices against quality attribute requirements (performance, availability, security, modifiability) to expose risk areas and tradeoffs.

Operations & Business Analysis Frameworks

• SIPOC (Suppliers, Inputs, Process, Outputs, Customers)

• Best For: Process optimization and DevOps workflow mapping.

• Focus: A Six Sigma tool that maps high-level boundaries of an end-to-end integration or system workflow to identify where bottlenecks and dependencies occur.

• CATWOE (Clients, Actors, Transformation, Worldview, Owner, Environmental constraints)

• Best For: Business Analysis (BA) root-cause discovery.

• Focus: Analyzes business problems by examining the broader system ecosystem and all human or technical stakeholders impacted by a proposed change.

Functional Requirements & Specifications Frameworks

• INVEST (Independent, Negotiable, Valuable, Estimable, Small, Testable)

• Best For: Agile backlog refinement and user story design.

• Focus: A checklist used to evaluate the quality of a requirement before development starts.

• Gherkin / BDD (Given, When, Then)

• Best For: Technical BA specifications and acceptance testing.

• Focus: Translates functional logic into readable scenarios: Given a initial state, When an integration action occurs, Then verify the specific outcome.