• 00:00 Welcome, and what's coming up
• 00:56 Building charity dashboards from publicly available data
• 02:11 The free UK government API catalogue
• 03:43 The national fish population database, and a pun Anna immediately regrets
• 04:34 Craig's first email response bot, and why a human still checks every reply
• 05:34 Voice-to-voice translation in about 90 languages, and where it still trips up
• 06:41 Disclosing you're talking to a bot, and why bots earned their bad reputation
• 07:38 When AI can't understand a Yorkshire accent
• 08:15 AI agents you can delegate to, and why IT teams get nervous
• 11:05 What agents doing junior admin work mean for the jobs pipeline
• 12:12 Oops AI Did It Again: AI-generated charity appeals and the trust problem
• 13:59 Why giving is drifting from institutions to individual faces online
• 18:12 Regulation, and not throwing the baby out with the bathwater
• 18:48 Coming up: the big AI debate

Mentioned in this episode

• UK government API Catalogue. The public sector API directory Craig recommends, covering DWP, DVLA, the Office for National Statistics, the British Geological Survey and more: api.gov.uk. Each API carries its own licensing and access terms, so check before you build.
• The AI GP receptionist that struggles with Yorkshire accents. Craig said on air he'd dig this out for the show notes. Healthwatch Rotherham raised concerns that EMMA, an AI phone receptionist used by GP surgeries, could not reliably understand broad Yorkshire accents or speech affected by an impediment, with some patients giving up and walking in instead: Silicon UK. It was GP surgeries rather than a council.
• Make. The automation platform Craig and Anna both use: make.com. Its AI agents were announced in April 2025, and its plain-English building assistant is called Maia. There is a generous free non-profit plan.
• The Lily Jay Foundation investigation. ABC News Verify in Australia reported finding AI-generated elements in charity videos posted by influencer Lily Jay, and could not verify the orphanage in Uganda: ABC News Verify. The foundation has since responded, saying AI appeared only in a small number of introductory clips and denying that it fabricated projects or aid deliveries: The New Arab.

• 00:00 Welcome
• 01:40 What we've been building: Anna's holiday vibe coding and a homemade family planner app
• 04:34 Craig's fundraising brain: 20 years of articles, talks and his own book, made searchable
• 09:17 Back to basics: companies, interfaces and models
• 10:06 The chef who read every cookbook but never tasted food
• 11:44 Why the model matters, and the exponential curve in expert hours
• 14:04 The otter test & Will Smith eating spaghetti
• 15:13 Tokens, context windows, and why five Harry Potter books is a useful yardstick
• 16:48 What are your tokens actually costing you? Nobody quite knows
• 19:14 Full disclosure: we're AI enthusiasts, but we're not blind to the problems
• 19:53 Oops AI Did It Again: Deloitte and KPMG get caught not checking their AI
• 21:44 The new EU AI Act rules: watermarking, manipulation and the social scoring ban
• 24:44 Why the sector's GDPR overcorrection is the thing to avoid
• 26:58 Meet the AI vegan
• 28:28 What's coming next: AI ethics and charity AI policies

Mentioned in this episode

• Ethan Mollick's "otter on a plane using wifi" test. Mollick's own account of how a silly prompt became his benchmark for AI image progress: The recent history of AI in 32 otters.
• The "Will Smith eating spaghetti" test. How a broken 2023 clip became the informal yardstick for AI video: Wikipedia, and TechCrunch on the odd AI benchmarks that took off.
• Deloitte Australia's AI-generated report. Deloitte partially refunded a A$440,000 report after fabricated references and a made-up court quote were found: Associated Press coverage. The refund was later reported at about A$97,000: CFO Dive.
• KPMG's withdrawn agentic AI report. "Redefining excellence in the age of agentic AI" was pulled after numerous the case studies were found to be invented: International Accounting Bulletin, Storyboard18.
• Obsidian, the note-linking app Craig recreated the idea of with Claude: obsidian.md.
• Get in touch: beyondtheslop@gmail.com, or find Craig Linton and Anna Hessenbruch on LinkedIn.

• 00:00 Who we are, and why two fundraisers started an AI podcast
• 01:22 How Craig and Anna met, and the supporter experience thread that led here
• 02:45 Why data and CRMs have always been the barrier, and what changes now
• 03:12 A data project that took six months and tens of thousands of pounds, rebuilt in half a day
• 04:11 What "slop" actually means, and why the show is named after it
• 06:14 Why writing tasks are only a small slice of what AI does
• 07:10 Oops AI Did It Again: the professor, the invisible text, and the random fact about Malaysia
• 08:31 The real tension: hurry up with AI, but don't you dare get it wrong
• 09:21 The Britney drinking game, and where the show nearly got its name
• 11:13 The regular segments: what we're building, the AI news that matters, and tools we actually use
• 15:08 Coming up: AI ethics, choosing use cases, and a plain-English guide to the different AI models
• 16:41 Next up: Craig and Anna debate whether AI is a force for good or a destructive menace