menno.sh

draw.io if it was good

2026-04-25T00:00:00+00:00

</p>

I found mockdown.design</a> on Marcin Wichary's blog</a> and it makes me want to replace draw.io -- or worse, excalidraw -- with this for everything. In any case, I drew a cool sail boat (or tropical island).</p>

      ┌──────────────┐
      │ sailboat.txt │
      └──────────────┘



                       ░░░░░░░░
                 ░░░░░▓▓▓▓▓▓▓▓ ░░░░░░
                ▓▓▓▓▓████████ ▓▓▓▓▓▓░░░░
               █████▓▓▓▓▓▓▓▓ ██████▓▓▓▓░░░
              ▓▓▓▓▓░░░░░░░░█▓▓▓▓▓▓████▓▓▓░░
             ░░░░░░░       █░░░░░▓▓▓▓███▓▓
                           █    ░░░░▓▓▓██
                           █       ░░░▓▓
                           █         ░░
            ░░░░░░░░░░     █                 ░░░░
           ▓▓▓▓▓▓▓▓▓▓░░░░░░█░░░░░░░░░░░░░░░░▓▓▓▓
          ██████████▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓████
         ▓▓▓▓▓▓▓▓▓▓███████████████████████▓▓▓▓
        ░░░░░░░░░░▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓░░░░
                 ░░░░░░░░░░░░░░░░░░░░░░░░░
</pre>

Weekoverzicht 2026-04-24

2026-04-24T00:00:00+00:00

Storylines</h2>

Antonelli leidt het kampioenschap als jongste WK-leider ooit. Mercedes helpt hem nuchter te blijven met grappen in de garage, maar de Italiaan weet dat de vijf weken break tussen Japan en Miami zijn momentum kan breken.</li>
Red Bull bevestigt dat de power unit niet wordt geüpgraded ondanks achterstand op de concurrentie. Hadjar maakt zijn promotie naar het hoofdteam intussen waar — al noemt hij het tweede zitje openhartig "weird" om vooraf naartoe te werken.</li>
De FIA en F1 hebben tussentijdse regelwijzigingen rond energy-management afgesproken voor Miami. Doel: kwalificaties weer scherper op de limiet. Of het ook werkt? Coulthard waarschuwt dat Miami nog niet het volledige beeld zal geven.</li> </ul>
Nieuws</h2>
Antonelli wil momentum vasthouden</strong> — De WK-leider verloor het meest door de vijf weken pauze, maar werkte op de simulator om scherp te blijven. Bron: autosport</a></p>
Red Bull bevriest power unit</strong> — Geen upgrade ondanks ADUO-druk, bevestigde het team. Bron: reddit</a></p>
F1-regels aangepast voor Miami</strong> — FIA en teams kwamen overeen op energy-management-tweaks om kwalificaties weer op de limiet te krijgen. Bron: motorsport</a></p>
Hadjar over Red Bull-zitje</strong> — De Frans-Algerijnse rookie bekent dat hij vooraf zorgen had over het "weird" tweede zitje, maar dat de promotie altijd zijn plan was. Bron: motorsport</a></p>
Coulthard: Miami onthult nog niet alles</strong> — De ex-coureur tempert verwachtingen over de impact van de nieuwe regelwijzigingen op energy-management. Bron: motorsport</a></p>
Stand na 3 races</h2>
Coureurs</h3>
#</th> Coureur</th> Team</th> Ptn</th> Wins</th></tr></thead>

1</td> Andrea Kimi Antonelli</td> Mercedes</td> 72</td> 2</td></tr>
2</td> George Russell</td> Mercedes</td> 63</td> 1</td></tr>
3</td> Charles Leclerc</td> Ferrari</td> 49</td> 0</td></tr> </tbody></table>
Constructors</h3>
#</th> Team</th> Ptn</th> Wins</th></tr></thead>

1</td> Mercedes</td> 135</td> 3</td></tr>
2</td> Ferrari</td> 90</td> 0</td></tr>
3</td> McLaren</td> 46</td> 0</td></tr> </tbody></table>
Komend weekend — Miami GP (Miami Gardens)</h2>
Sessies:</p>
Dag</th> Datum</th> Tijd CEST</th> Sessie</th></tr></thead>

Vr</td> 2026-05-01</td> 18:30</td> Practice 1</td></tr>
Vr</td> 2026-05-01</td> 22:30</td> Sprint Qualifying</td></tr>
Za</td> 2026-05-02</td> 18:00</td> Sprint</td></tr>
Za</td> 2026-05-02</td> 22:00</td> Qualifying</td></tr>
Zo</td> 2026-05-03</td> 22:00</td> Race</td></tr> </tbody></table>

Proost op Heineken als hoofdsponsor: op naar nog 750 jaar belastingontwijking en oorlogsfinanciering

2024-04-25T00:00:00+00:00

Opiniestuk gepubliceerd in Het Parool</a>.</p>

Take some time to think about time

2022-06-22T00:00:00+00:00

Originally published on Medium</a>.</p>
Y2K, archaic databases and spoOoOky integer overflows</em></p>
</p>
At the moment of writing it is the 22nd of June, the day after summer solstice. The pessimist in me laments that the days will only get shorter from now on; the optimist knows that this also heralds the beginning of summer, meaning the best should be yet to come.</p>
In the CS paradigm, time has often been a concept that proves more difficult to define than you might initially think. Some of you may have consciously experienced Y2K: because time was traditionally encoded in MMDDYY format, the whole (digital) world was going to go up in flames with the jump from '99 to '00.</p>
There is a misconception that nothing was going to go wrong in the end and systems were able to deal with this just fine, but this forgoes the fact that many software engineers at the time worked around the clock to mitigate the damage.</p>
Today, Unix-based systems run on Unix time</em>, defined as the number of seconds since 00:00:00 UTC on January 1, 1970. This method, however, also has an Achilles' heel. The convention was to store this value as a signed 32-bit integer, which means that at 2³²-1 — or 03:14:07 UTC on January 19, 2038 — we are facing an integer overflow. The time will jump to -2³², better known as Friday the 13th (!) of December 1901. Systems that are particularly susceptible to this are, for example, SQL databases that store time as INT32.</p>
The takeaway? If you're (still) in the data field at the end of the fourth decade of this century — and we're still using SQL databases instead of some blockchain or quantum-superpositioned and entangled qubits — take some time</em> to check whether you have INT32 columns that contain time data. If so, simply converting them to INT64 should alleviate the issue for another 292 billion years, or 584 billion solstices.</p>

Interpretation of Deep Learning Models in Bearing Fault Diagnosis

2021-11-24T00:00:00+00:00

Paper published at the Annual Conference of the PHM Society 2021, together with Cees Taal, Sebastián Echeverri Restrepo, and Alireza Azarfar (all SKF).</p>
Abstract</h2>
In recent years, data-driven techniques such as deep learning (DL) have been widely represented in the literature in the field of bearing vibration condition monitoring. While these approaches achieve excellent performance in classifying bearing faults on controlled laboratory data sets, there is little information available about their applicability to more realistic working conditions.</p>
As a first step towards revealing the generalizability of DL models, we aim to understand the underlying representations that DL networks use to classify bearing defects. An interpretable DL model can give us hints on how to increase its transferability by, e.g., using data augmentation, changing input representations and/or adapting model architectures. We use the Grad-CAM algorithm along with signal transformations to identify the elements of the input spectrogram that contribute to class attribution.</p>
The results show that removing time-domain information from the spectrogram has a minor impact on its performance. Instead, the network learns distinct average frequency profiles. We therefore conclude that the networks learn signal features very specific to the physical properties of the specific test setup, such as the frequency response function, rather than more general features related to bearing defects.</p>
Full paper (PDF)</a> · DOI: 10.36001/phmconf.2021.v13i1.3047</a></p>