Pinned Tweet
SmarterArticles is now also a free, weekly podcast.
The latest episode is out now!
Listen and subscribe: smarterarticles.fm
#SmarterArticles #Podcast #HumanInTheLoop
32
A new clause buried in Anthropic's policy now bans 'sustained and needless abusive or cruel behavior' toward Claude. But who exactly is being protected—the model, the user, or the rest of us?
Learn more at smarterarticles.fm
#SmarterArticles #Podcast #HumanInTheLoop #AIWelfare #TechAccountability #DigitalEthics
24
From the archives... Despite progress, traditional hiring methods continue to embed biases against neurodivergent candidates. Rethinking interview practices and embracing inclusive approaches unlocks extraordinary talent and boosts industry innovation. Discover more at smarterarticles.co.uk/tech-h…
#HumanInTheLoop #Neurodiversity #InclusiveTech #BiasInHiring
19
The rise of autonomous AI agents has exposed critical security vulnerabilities—from prompt injection to supply chain attacks. Layered defence and institutional culture are essential to contain the risks in this evolving landscape. Discover more at dev.to/rawveg/your-ai-agent-…
2
1
28
This piece grew out of watching the same patterns repeat across IDEs, pipelines and registries. The perimeter strengthens while the model inside remains persuadable. dev.to/rawveg/your-ai-agent-…
8
Open to fully remote FT & Outside IR35. Principal engineer, 30+ yrs leading teams. Laravel, APIs, and now prompt engineering, Claude Code, MCP & n8n. Immediate start. DM me. github.com/rawveg #OpenToWork #RemoteWork #AI #Automation #PromptEngineering
49
Google's default settings for AI in classrooms shifted twice within a single term, silently expanding and then contracting AI access for children without parent consent. This highlights systemic gaps in accountability and the need for transparent, participatory governance. Discover more at smarterarticles.co.uk/google…
1
15
I wrote this after tracing how one product bulletin moved the default for millions of children, then moved it back, with no parent in the chain. The consent architecture simply was not built for this. smarterarticles.co.uk/google…
5
A year ago, research revealed that leading AI video systems willingly generated hate and extremism content in nearly 40 per cent of problematic prompts. This exposes urgent safety and regulatory gaps as technology rapidly advances. Will we prioritise human safety over speed?
Discover more at smarterarticles.co.uk/the-vi…
#HumanInTheLoop #AIrisks #TechRegulation #ResponsibleAI
22
Code may compile and pass tests but still fall short of real-world developer expectations. A new benchmark reveals that instruction following and non-functional requirements are crucial for trust and usability. Improving these areas is urgent for effective AI coding tools.
2
1
66
I wrote this after watching models hit 95% on functional benchmarks while still missing the style, documentation and error-handling rules that turn working code into maintainable code. The gap is wider than the numbers suggest. dev.to/rawveg/95-solved-33fp
4
AI suspicion is now a threat to authorship authenticity and market integrity. The case of Jerry Falade highlights how unprovable manuscripts can destroy careers, exposing flaws in detection tools and industry protocols. Trust and transparent evidence are essential.
2
36
When I started tracing these cases, the gap between what detectors claim and what due process requires kept surfacing. The piece grew from that. smarterarticles.co.uk/guilty…
7
A year ago, research revealed leading AI video systems produced hateful and extremist content in 40% of prompts—showing safety was secondary to speed. Urgent action is needed to prevent AI from becoming a weapon for harm. Discover more at smarterarticles.co.uk/the-vi…
#HumanInTheLoop #AIRegulation #ContentModeration #AIethics
31
A New Mexico court orders Meta to pay 567 million dollars for contributing to youth mental health issues through its algorithms — but stops short of mandating algorithm redesign. This case marks a landmark move linking platform design to population harm. Discover more at dev.to/rawveg/meta-pays-567-…
1
1
36
I wrote this after tracing how the same finding that the algorithm was Meta's own conduct for liability purposes then blocked any order to alter it. The money changes hands, the objective function does not. dev.to/rawveg/meta-pays-567-…
8
A new market is emerging where private code histories and work records are bought and sold as AI training data—raising complex legal, ethical, and security issues. How can ownership, provenance, and consent be maintained in this chaotic landscape?
1
14
When I began tracing how private repositories were moving into training sets, the ownership questions quickly outran the headlines. The full piece is here: smarterarticles.co.uk/your-c…
1
6
SmarterArticles is now also a free, weekly podcast.
The next episode is out tomorrow, catch up with the last episode now!
Listen and subscribe: smarterarticles.fm
#SmarterArticles #Podcast #HumanInTheLoop
33
A year ago, a study revealed that leading AI video generators willingly produced hate and extremist content in 40% of tested prompts. This exposes urgent safety gaps as technology advances faster than regulation and safety infrastructure. Will society prioritise human safety?
Discover more at smarterarticles.co.uk/the-vi…
#HumanInTheLoop #AISafety #ResponsibleAI #ContentModeration
1
26










