Deborah Barnard

Deborah Barnard

@deborahwrites.co.uk

💻 Technical writer 🐱 Human to one loud cat https://deborahwrites.co.uk/

912 followers1.5K following329 posts

View full thread

The topics are also covered in the README - this isn't a replacement for a docs page/site. But it means folks who don't read the docs and go straight to your issues have another chance of seeing the info.

Feeling slightly bored after the end of the World Cup, so consoling myself by deciding to follow the Nations League. We're nearing an era of perpetual football: a major tournament every year (alternating men's and women's) and now the men's Nations League to fill some of the winter.

Deborah Barnard reposted

Replying to @eliothiggins.bsky.social

LLMs give people with delusional belief systems a way to organise those beliefs. The model fills in gaps, supplies connective tissue, and wraps the madness in formal language. It doesn't make the claims more credible to us, it just tidies them up.

2/ One interesting item from the research he cites is the vastly different performance of different models (TL;DR - Anthropic are doing a heck of a lot better at AI safety than their main competitors)

3/ "If it feels too perfectly tailored to me, if it presses my buttons, if it precisely tickles me where I like to be tickled or hurts me where I am vulnerable to being hurt, that’s a sign to check the facts." - this is a really useful tip, however hard it may be to practice.

4/ "What [institutions] would various rulers like to break up and destroy because they represent alternative sources of authority to their own narrative, and also there’s money to be made?" - Especially topical at the moment. The BBC and the Guardian both make mistakes, but are worth protecting.

2/ It's an interesting historical contextualisation for our current age, which I find reassuring. Humanity has done some awful things, but overall we've survived and improved. We might make it through this crisis too.

Deborah Barnard reposted

Zensical – a modern static site generator. We're excited to release @zensical.org, our new SSG that is compatible with Material for MkDocs: – 5x faster rebuilds – Modern design – Blazing-fast search Read the full announcement on our blog: squidfunk.github.io/mkdocs-mater... A thread ⬇ 1/7

squidfunk.github.ioZensical - A modern static site generator - Material for MkDocsWe are thrilled to announce Zensical, our next-gen static site generator that addresses and overcomes the technical limitations of MkDocs

2/ While I'm kinda sad to see my favourite tooling heading (slowly) towards the exit, I'm excited about the new and shiny, and very pleased to see new docs tooling being built by a team that's earned trust over a lot of years.

Deborah Barnard reposted

Back in the day, Twitter used to be really good if I set up a thread to connect freelancers with clients. Let's see if Bluesky can do it. Clients: if you're looking for freelancers/contractors, get in the comments Freelancers/contractors: get in comments Everyone else: boosts appreciated

Deborah Barnard reposted

we've already sold 1000 copies of The Secret Rules of the Terminal since it came out on Tuesday!! so to celebrate I'm giving away 1000 copies to anyone who can't afford to buy one. You can use code BUYONEGIVEONE to get a free PDF copy if $12 USD is a lot for you. wizardzines.com/zines/terminal

1/ Could we all solemnly swear not to be too judgemental of each other about AI? Because I've got a horrible feeling I know what the next impossible hiring trend is going to be: you'll need enough AI on your CV to get past the pro-AI people, but not so much that you deter the anti-AI people.

2/ The fun thing about being on both LinkedIn and Bluesky is I get to read the LinkedIn posts from companies pushing AI usage hard, and the Bluesky posts from people who've decided anyone who even touches AI is signalling complete incompetence.

Deborah Barnard reposted

If you were too young in 2003 here is what Bush told the American people: 1) The Iraq War won’t cost a lot of money 3) We won’t need a lot of troops 4) We’ll bring peace and freedom to Iraq 5) We’ll find WMDs 6) We’ll be welcomed as liberators 7) It will be easy 8) It won’t take long All lies.

1/ A surprising amount of government admin. My last passport renewal was quick and easy. So is setting up as a sole trader. Tax returns are probably as painless as they realistically can be. And the fact most people don't even have to do tax returns is nice.

This post is unavailable.
View full thread

7/ Consistent backing for Ukraine. Another thing that has managed to cross political lines (except Farage - and it's one of the things that could trip him up if he talks about it too much)

1/ A surprising amount of government admin. My last passport renewal was quick and easy. So is setting up as a sole trader. Tax returns are probably as painless as they realistically can be. And the fact most people don't even have to do tax returns is nice.

This post is unavailable.
View full thread

4/ The sense of humour really is a thing, and perhaps helps cross political borders in a way some countries just don't seem able to. I've yet to encounter anyone, even online, who doesn't laugh about the lettuce. And the fact Yes Minister is still funny and applicable to any government is something.

1/ A surprising amount of government admin. My last passport renewal was quick and easy. So is setting up as a sole trader. Tax returns are probably as painless as they realistically can be. And the fact most people don't even have to do tax returns is nice.

This post is unavailable.
View full thread
View full thread

2/ TL;DR: One of the positives of AI is that good writing for AI is mostly also good writing for humans (so I live in hope of companies increasingly prioritising documentation quality). Learn to be a good tech writer for your users, and you'll probably do a good job for your AI audience as well.

Deborah Barnard reposted

A lot of people say generative AI shouldn't infringe on copyright. These researchers actually tried to do it. The result: an 8 terabyte dataset of text that's openly licensed or in the public domain & 7 B parameter model that performs as well as Meta's Llama 7B www.washingtonpost.com/politics/202...

washingtonpost.comAnalysis | AI firms say they can’t respect copyright. These researchers tried.A new effort using only openly licensed data may have implications on thorny policy disputes around copyright and AI