Gareth Dwyer 🇳🇱🇨🇭🇿🇦
@sixhobbits.bsky.social
Bootstrapping Ritza.co - a technical writing agency. Follow me for slightly cynical takes on tech and marketing.
I mainly use BlueSky to post about Eskom data and loadshedding. See Twitter.com/sixhobbits for startup/tech stuff.
I mainly use BlueSky to post about Eskom data and loadshedding. See Twitter.com/sixhobbits for startup/tech stuff.
Eskom data - October 2025
Not anything too interesting this month - EAF down again due to an increase in unplanned outages, but still 'ok' overall.
Not anything too interesting this month - EAF down again due to an increase in unplanned outages, but still 'ok' overall.
November 4, 2025 at 6:54 AM
Eskom data - October 2025
Not anything too interesting this month - EAF down again due to an increase in unplanned outages, but still 'ok' overall.
Not anything too interesting this month - EAF down again due to an increase in unplanned outages, but still 'ok' overall.
Eskom data - September 2025
Another dip in unplanned outages takes us close to 2020 levels!
Another dip in unplanned outages takes us close to 2020 levels!
October 3, 2025 at 7:20 AM
Eskom data - September 2025
Another dip in unplanned outages takes us close to 2020 levels!
Another dip in unplanned outages takes us close to 2020 levels!
Eskom data - August 2025 (finally)
I finally got a reply to my data request for August data (usually takes a few hours / days, this time several weeks)
Metrics looking good compared to the last few months, but not that far off where we were a year ago.
Very nice dip in unplanned outages.
I finally got a reply to my data request for August data (usually takes a few hours / days, this time several weeks)
Metrics looking good compared to the last few months, but not that far off where we were a year ago.
Very nice dip in unplanned outages.
September 15, 2025 at 11:23 AM
Eskom data - August 2025 (finally)
I finally got a reply to my data request for August data (usually takes a few hours / days, this time several weeks)
Metrics looking good compared to the last few months, but not that far off where we were a year ago.
Very nice dip in unplanned outages.
I finally got a reply to my data request for August data (usually takes a few hours / days, this time several weeks)
Metrics looking good compared to the last few months, but not that far off where we were a year ago.
Very nice dip in unplanned outages.
Eskom data - July 2025
Sorry I missed June, I was travelling.
July is looking positive! Big decrease in total loss, and this time it's in the good way (decrease in unplanned, increase in planned)
Sorry I missed June, I was travelling.
July is looking positive! Big decrease in total loss, and this time it's in the good way (decrease in unplanned, increase in planned)
August 4, 2025 at 10:40 AM
Eskom data - July 2025
Sorry I missed June, I was travelling.
July is looking positive! Big decrease in total loss, and this time it's in the good way (decrease in unplanned, increase in planned)
Sorry I missed June, I was travelling.
July is looking positive! Big decrease in total loss, and this time it's in the good way (decrease in unplanned, increase in planned)
Starting a new thread to dump random stuff about Claude Code, mainly as notes for myself by maybe interesting to others too. 🧵
July 21, 2025 at 10:58 AM
Starting a new thread to dump random stuff about Claude Code, mainly as notes for myself by maybe interesting to others too. 🧵
Eskom data - May 2025
- Unplanned outages still looking bad
- Small bump from new wind installations
- Diesel use trending down again
🧵
- Unplanned outages still looking bad
- Small bump from new wind installations
- Diesel use trending down again
🧵
June 3, 2025 at 12:27 PM
Eskom data - May 2025
- Unplanned outages still looking bad
- Small bump from new wind installations
- Diesel use trending down again
🧵
- Unplanned outages still looking bad
- Small bump from new wind installations
- Diesel use trending down again
🧵
Eskom April data 🔋🔌💡
Note: the sqlite databases that all the metabase queries are based on are now available for direct download at unofficialeskom.com/dashboards/#.... I still have a TODO to create a data dictionary and explain the columns etc, but it's pretty straight forward.
🧵
Note: the sqlite databases that all the metabase queries are based on are now available for direct download at unofficialeskom.com/dashboards/#.... I still have a TODO to create a data dictionary and explain the columns etc, but it's pretty straight forward.
🧵
May 8, 2025 at 12:18 PM
Eskom April data 🔋🔌💡
Note: the sqlite databases that all the metabase queries are based on are now available for direct download at unofficialeskom.com/dashboards/#.... I still have a TODO to create a data dictionary and explain the columns etc, but it's pretty straight forward.
🧵
Note: the sqlite databases that all the metabase queries are based on are now available for direct download at unofficialeskom.com/dashboards/#.... I still have a TODO to create a data dictionary and explain the columns etc, but it's pretty straight forward.
🧵
Fact checking Eskom's winter outlook claims
(x.com/Eskom_SA/sta...)
Eskom is great at using data to imply that things are going well, even when the data actually shows that things are not going well.
They do this by cherry-picking data points and not precisely defining the time period. 🧵
(x.com/Eskom_SA/sta...)
Eskom is great at using data to imply that things are going well, even when the data actually shows that things are not going well.
They do this by cherry-picking data points and not precisely defining the time period. 🧵
May 5, 2025 at 10:26 AM
Fact checking Eskom's winter outlook claims
(x.com/Eskom_SA/sta...)
Eskom is great at using data to imply that things are going well, even when the data actually shows that things are not going well.
They do this by cherry-picking data points and not precisely defining the time period. 🧵
(x.com/Eskom_SA/sta...)
Eskom is great at using data to imply that things are going well, even when the data actually shows that things are not going well.
They do this by cherry-picking data points and not precisely defining the time period. 🧵
Eskom data from March 2025
- Planned outages down (= bad, less maintenance) and unplanned outages up (= bad, more unexpected failings). But overall leaves us with a similar EAF (availability factor)
🧵
- Planned outages down (= bad, less maintenance) and unplanned outages up (= bad, more unexpected failings). But overall leaves us with a similar EAF (availability factor)
🧵
April 1, 2025 at 1:30 PM
Eskom data from March 2025
- Planned outages down (= bad, less maintenance) and unplanned outages up (= bad, more unexpected failings). But overall leaves us with a similar EAF (availability factor)
🧵
- Planned outages down (= bad, less maintenance) and unplanned outages up (= bad, more unexpected failings). But overall leaves us with a similar EAF (availability factor)
🧵
New Eskom (South African electricity) data from February.
- Planned up, unplanned down = good
- EAF (energy availability factor) slightly better to 57.5%, far off the March 2025 70% target that we were 'on track' to meet in November 2024.
- Planned up, unplanned down = good
- EAF (energy availability factor) slightly better to 57.5%, far off the March 2025 70% target that we were 'on track' to meet in November 2024.
March 3, 2025 at 6:42 AM
New Eskom (South African electricity) data from February.
- Planned up, unplanned down = good
- EAF (energy availability factor) slightly better to 57.5%, far off the March 2025 70% target that we were 'on track' to meet in November 2024.
- Planned up, unplanned down = good
- EAF (energy availability factor) slightly better to 57.5%, far off the March 2025 70% target that we were 'on track' to meet in November 2024.
So we got loadshedding again in South Africa for the first time in 10 months. Some people think it's suspicious that it happened exactly after NERSA (the energy regulator) denied the price increase that Eskom (the producer) was asking for. Some say it's a coincidence.
Let's look at some data 🧵
Let's look at some data 🧵
February 3, 2025 at 10:38 AM
So we got loadshedding again in South Africa for the first time in 10 months. Some people think it's suspicious that it happened exactly after NERSA (the energy regulator) denied the price increase that Eskom (the producer) was asking for. Some say it's a coincidence.
Let's look at some data 🧵
Let's look at some data 🧵
I use LLMs every day, and yet I still find it super hard to predict what they will get right and what will go horribly wrong.
It's hard to fight the hindsight bias because after it works or doesn't work I always _feel_ like I knew, but if I am strict and make an actual prediction, I'm often wrong.
It's hard to fight the hindsight bias because after it works or doesn't work I always _feel_ like I knew, but if I am strict and make an actual prediction, I'm often wrong.
January 24, 2025 at 1:51 PM
I use LLMs every day, and yet I still find it super hard to predict what they will get right and what will go horribly wrong.
It's hard to fight the hindsight bias because after it works or doesn't work I always _feel_ like I knew, but if I am strict and make an actual prediction, I'm often wrong.
It's hard to fight the hindsight bias because after it works or doesn't work I always _feel_ like I knew, but if I am strict and make an actual prediction, I'm often wrong.
Reposted by Gareth Dwyer 🇳🇱🇨🇭🇿🇦
If you have a browser that supports WebGPU (like Google Chrome) you try out the DeepSeek-R1 model based on Qwen2.5-Math-1.5B directly in your browser!
It's a 1.28GB page load: huggingface.co/spaces/webml...
It's a 1.28GB page load: huggingface.co/spaces/webml...
January 21, 2025 at 10:06 PM
If you have a browser that supports WebGPU (like Google Chrome) you try out the DeepSeek-R1 model based on Qwen2.5-Math-1.5B directly in your browser!
It's a 1.28GB page load: huggingface.co/spaces/webml...
It's a 1.28GB page load: huggingface.co/spaces/webml...
I asked Eskom for clarity about the Koeberg schedule, as I previously read that Unit 1 had more scheduled downtime after the return of Unit 2, but the latest media releases didn't mention it again.
They confirmed that Unit 1 is scheduled for more downtime in Feb, so no net increase in nuclear yet.
They confirmed that Unit 1 is scheduled for more downtime in Feb, so no net increase in nuclear yet.
January 16, 2025 at 8:47 AM
I asked Eskom for clarity about the Koeberg schedule, as I previously read that Unit 1 had more scheduled downtime after the return of Unit 2, but the latest media releases didn't mention it again.
They confirmed that Unit 1 is scheduled for more downtime in Feb, so no net increase in nuclear yet.
They confirmed that Unit 1 is scheduled for more downtime in Feb, so no net increase in nuclear yet.
Reposted by Gareth Dwyer 🇳🇱🇨🇭🇿🇦
Huh oh. As a committed anti-conspiracy-theorist for the "phones listen to your microphone and targeted ads based on what you say" thing I can tell this is going to be exhausting
I think probably 95% of speculation that your phone is secretly listening to you is overblown, but man the tech industry really does not do itself any favors with that 5% arstechnica.com/tech-policy/...
Siri “unintentionally” recorded private convos; Apple agrees to pay $95M
Apple users may get $20 each for up to five Siri-enabled devices.
arstechnica.com
January 2, 2025 at 9:19 PM
Huh oh. As a committed anti-conspiracy-theorist for the "phones listen to your microphone and targeted ads based on what you say" thing I can tell this is going to be exhausting
Not much in the December data for Eskom.
- Increase in planned (normal) and unplanned (unfortunate) maintenance led to lower EAF
- As a result, more running of OCGT emergency generators - more than Oct/Nov, slightly less than Sep
-
- Increase in planned (normal) and unplanned (unfortunate) maintenance led to lower EAF
- As a result, more running of OCGT emergency generators - more than Oct/Nov, slightly less than Sep
-
January 3, 2025 at 6:56 AM
Not much in the December data for Eskom.
- Increase in planned (normal) and unplanned (unfortunate) maintenance led to lower EAF
- As a result, more running of OCGT emergency generators - more than Oct/Nov, slightly less than Sep
-
- Increase in planned (normal) and unplanned (unfortunate) maintenance led to lower EAF
- As a result, more running of OCGT emergency generators - more than Oct/Nov, slightly less than Sep
-
this is interesting. I remember being surprised when LLMs suddenly could do accurate arithmetic.
I reproduced similar cases with GPT and Claude now - if you ask maths directly it works. Indirect doesn't.
@simonwillison.net any take on this?
I reproduced similar cases with GPT and Claude now - if you ask maths directly it works. Indirect doesn't.
@simonwillison.net any take on this?
January 2, 2025 at 9:18 AM
this is interesting. I remember being surprised when LLMs suddenly could do accurate arithmetic.
I reproduced similar cases with GPT and Claude now - if you ask maths directly it works. Indirect doesn't.
@simonwillison.net any take on this?
I reproduced similar cases with GPT and Claude now - if you ask maths directly it works. Indirect doesn't.
@simonwillison.net any take on this?
I've always been on the fence about whether Telegram is a group of cool engineers building a great product with amazing UX because they want to, or if it's a huge datagrab operation for <reasons>.
The fact that I get this pop-up so often definitely pushes me towards the second hypothesis.
The fact that I get this pop-up so often definitely pushes me towards the second hypothesis.
December 11, 2024 at 9:34 AM
I've always been on the fence about whether Telegram is a group of cool engineers building a great product with amazing UX because they want to, or if it's a huge datagrab operation for <reasons>.
The fact that I get this pop-up so often definitely pushes me towards the second hypothesis.
The fact that I get this pop-up so often definitely pushes me towards the second hypothesis.
Two biggest mistakes to make in entrepreneurship
- Give up / pivot too soon
- Don't give up / pivot soon enough
The fun part is you never know which one you're making at any given time.
- Give up / pivot too soon
- Don't give up / pivot soon enough
The fun part is you never know which one you're making at any given time.
December 11, 2024 at 7:13 AM
Two biggest mistakes to make in entrepreneurship
- Give up / pivot too soon
- Don't give up / pivot soon enough
The fun part is you never know which one you're making at any given time.
- Give up / pivot too soon
- Don't give up / pivot soon enough
The fun part is you never know which one you're making at any given time.
Looked at some longer term data and trends for 2024, had a very clear picture of what to do in 2025, including some difficult decisions and changes.
Went for a walk and did a complete 180 on all of it.
Founders/indie people, how often do you wish you just had someone tell you what to do again?
Went for a walk and did a complete 180 on all of it.
Founders/indie people, how often do you wish you just had someone tell you what to do again?
December 10, 2024 at 3:22 PM
Looked at some longer term data and trends for 2024, had a very clear picture of what to do in 2025, including some difficult decisions and changes.
Went for a walk and did a complete 180 on all of it.
Founders/indie people, how often do you wish you just had someone tell you what to do again?
Went for a walk and did a complete 180 on all of it.
Founders/indie people, how often do you wish you just had someone tell you what to do again?
If you thought you didn't achieve enough in the last 3 years, remember that Zendesk just sent me an email to brag that they got into Gartner Quadrant 2021. I thought it was a typo so I checked 2023, and nope they're not there any more.
December 9, 2024 at 12:01 PM
If you thought you didn't achieve enough in the last 3 years, remember that Zendesk just sent me an email to brag that they got into Gartner Quadrant 2021. I thought it was a typo so I checked 2023, and nope they're not there any more.
ok this is meta and an example of the thing it is advocating against, but how about everyone stops posting meta platform stuff?
I tried/am trying to move from reddit/twitter to mastodon/bluesky/lemmy and each time I open the new one I see only posts about mastodon/bluesky/lemmy, so I close the page
I tried/am trying to move from reddit/twitter to mastodon/bluesky/lemmy and each time I open the new one I see only posts about mastodon/bluesky/lemmy, so I close the page
December 3, 2024 at 8:31 AM
ok this is meta and an example of the thing it is advocating against, but how about everyone stops posting meta platform stuff?
I tried/am trying to move from reddit/twitter to mastodon/bluesky/lemmy and each time I open the new one I see only posts about mastodon/bluesky/lemmy, so I close the page
I tried/am trying to move from reddit/twitter to mastodon/bluesky/lemmy and each time I open the new one I see only posts about mastodon/bluesky/lemmy, so I close the page
Testing out bluesky with my monthly update on Eskom's data that I usually do on a Slack channel.
Eskom is South Africa's state-owned electricity provider. I started gathering and analysing the data when we had loadshedding (power outages) nearly every day.
Nov highlight: record export (avg 2GW)
Eskom is South Africa's state-owned electricity provider. I started gathering and analysing the data when we had loadshedding (power outages) nearly every day.
Nov highlight: record export (avg 2GW)
December 2, 2024 at 9:16 AM
Testing out bluesky with my monthly update on Eskom's data that I usually do on a Slack channel.
Eskom is South Africa's state-owned electricity provider. I started gathering and analysing the data when we had loadshedding (power outages) nearly every day.
Nov highlight: record export (avg 2GW)
Eskom is South Africa's state-owned electricity provider. I started gathering and analysing the data when we had loadshedding (power outages) nearly every day.
Nov highlight: record export (avg 2GW)