Andy David

You really need to be humanmaxxing

This week, on Intelligence Snacks,

and I chatted with

from Finite Computer about building agents that give people more room for the human stuff like exercising judgment, developing taste, finding meaningful problems and connecting with others.

Episode 75 of Intelligence Snacks is live.

and I got into Dead Internet Theory, why AI might push people towards smaller trusted networks, whether chat and documents are actually the right interfaces for working with agents, and why attention, judgement and agency are becoming the real bottlenecks as AI capability gets cheaper.

Intelligence Snacks now has a graph view to explore what

and I have been chatting about on the pod

Only a few eps uploaded so far but you can check it out here

https://intelligencesnacks.com/graph/

People are starting to realise you want more deterministic code + agents executing on parts of the process that genuinely need them, and less throwing frontier intelligence at everything.

Even more so if you're running a business. Companies over index on determinism because they want predictability.

This is why I'm so bullish on Wingman Apps, or WApps.

This is Snack Studio, it's a purpose-built WApp for distilling our podcast episodes and transcripts into a mixtape of publishable ideas and themes, AKA Intelligence Snacks.

Snack Studio feels like a normal web app, you can manage episodes, review generated content, select thumbnails, edit contributors and publish work.

Under the surface though, it's powered by Wingman.

Wingman provides the intelligence and automation layer behind the interface. When Snack Studio processes a transcript, drafts Snacks, generates artwork or publishes an episode, the WApp sends that work to Wingman as a pipeline.

Wingman coordinates the required agents and tools, tracks the work as it progresses, and sends the results back to the Snack Studio interface. The WApp updates as those results arrive, so I can stay focused on the content rather than manually operating the underlying automation.

A WApp can have its own interface, data model and rules while drawing on Wingman’s common capabilities underneath.

In other words, Snack Studio provides the specialised workflow and user experience. Wingman supplies the reusable infrastructure for intelligence, pipelines, permissions, execution and communication with external systems.

This is how I think most people will end up using AI, particularly in a business - codify existing business processes into highly personalised software that also takes that work out of chat.

If you missed Intelligence Snacks this week with Piers from

live on

, the episode is also now live on YouTube.

, Piers and I got into whether AI agents should have wallets, and what happens when they can earn, spend, rent infrastructure and act without asking a human to approve every step.

We also explored the collision between agents, Bitcoin, Nostr and sovereign infrastructure, agent payments, self-funded compute, abuse, identity, tax, cybersecurity and liability.

If you want the TLDR check out the latest snacks from ep 69 here: https://intelligencesnacks.com/episodes/episode-069/

We're

Great chat with

and

on the Intelligence Snacks podcast! Thanks for coming on Rod will need to do that again soon

If you missed Intelligence Snacks on

yesterday fear not, the full episode with

is now up on YouTube

We dug into the shift toward local AI models and what that means for privacy, sovereignty and the future of AI adoption, and why open-weight models are now capable enough for everyday knowledge work

We also explored why the next phase of AI will be defined by the orchestration layer that coordinates intelligence, rather than the models themselves, and how we're using Wingman in this way to run our business and help other SME's do the same

View from the Intelligence Snacks podcast studio this week featuring

and some guy having a mid life crisis in his convertible

YouTube out tomorrow as usual

First live stream of the rebranded pod - Intelligence Snacks - went off nicely. Might have to start streaming the pre and post pod exclusively on nostr 👀

YouTube drops tomorrow as per

https://shosho.live/live/naddr1qvzqqqrkvupzppwlsg4gvkvll05pg0d3u8sm7tgk97nql359ce23t937vl8awjvlqqjrsdtz8qmnjc3h94jxzdek956ryvp494skywrr95unxveex93n2wp5x5cny4d9j0c

We're live!

Great chat with

on the

this week about the idea that AI is doing for software development what power tools did for woodworking

We also talked about vibe coding, building your own agent harness, personal software, AI-generated technical debt, architecture, and why human attention still matters

39 --> 40

Excellent chat about models vs harnesses with

on the

this week.

We got into agent orchestration, context management, reasoning models, open source local models, inference economics, and the business strategies of OpenAI, Anthropic, Mistral, and the Chinese labs.

Check it out. Another banger episode with nostr:nprofile1qyvhwumn8ghj7urjv4kkjatd9ec8y6tdv9kzumn9wshsz9thwden5te0wfjkccte9ejxzmt4wvhxjme0qqs9dcwyxjc9rp2qayleajjygr9pvyfe34z4h8u3q2g3gsharw508kqt6e8hx

and I sat down with the one and only

on The

to chat about how AI is upending the billable hour in professional services.

Got into some awesome territory and covered things like:

➡️ Shawn's outer loop / inner loop framework - how to think about what to hand to AI and what to protect, with client relationships on one side, core IP on the other, and everything in between up for grabs.

➡️ The judgment sandwich - you need judgment to point the tools at the right problem and judgment to assess whether what came back is any good. AI handles the middle.

➡️ 1,000 true fans meets 10 agents - Shawn's reframe of Kevin Kelly's idea that you only need 1,000 real customers to have a real business. Now you also have 10 AI agents as your workforce. One person with the right clients and the right stack can do what used to require a small firm.

➡️ Enshittification - Shawn references Cory Doctorow's pattern where platforms first serve users, then extract from them. AI labs have the same structural incentive, which is why depending entirely on one of them is a trap.

➡️ The death of the commodity career - AI isn't just replacing tasks. It's replacing the career path that turned raw graduates into people with genuine judgment, through years of doing the work.

Definitely worth a listen!

Thanks for coming on nostr:nprofile1qyvhwumn8ghj7urjv4kkjatd9ec8y6tdv9kzumn9wshsz8rhwden5te0wfjkccte9e3xjarrda5kuurpwf4jucm0d5hsqgx8aknxpf4usfc9xr5zkjnhz2kdaghrrhq22mude926cqy7lktuscdh439v, already looking forward to the next one!

Enjoyed giving this episode with

from

a listen back from last week. Here's a snippet.

Another special guest on this weeks Big Episode 58 dropping tomorrow! 👀

Can anyone with a YouTube channel ELI5 how the YouTube algo works? I have no idea how we get two likes and six comments but still only 14 views

cool gear for people that quite like AI - coming soon 👀

soft launching so I actually finish this side quest

Idea Generation Machine courtesy of

Just dropped the latest episode of the

Pretty obvious by now that the barrier to making things has fallen dramatically, so we’ll get more stuff that probably didn’t need to exist.

But we’ll also get more genuinely useful niche things from people who deeply understand a problem or just want to see a thing exist in the world.

Was listening to DHH about building Omarchy and it feels like this is just such a good time to be a builder.

and I chatted about this on the pod this week

This week on the pod:

Agents are like humans, give them vague goals and they'll hit them surprisingly well, but that's not sustainable or efficient.

introduced "intelligence snacks" which are those small moments in otherwise deterministic workflows where you actually need AI to make a decision.

Most of what runs a business could just be scripts, but the snacks are where the magic happens.

Also we explored why Chamath is wrong when he questions the value of AI on the basis that enterprise hasn't seen widespread, positive margin expansion. Turns out enterprise is just the wrong place to look.

This week was our 1st Birthday on the

! Happy Birthday us!

and I chatted about why throwing agents at business problems is the mid-curve play, it's unpredictable, and likely to be undercut by anyone who takes the extra step to encode their business into software

Good chat

https://youtu.be/YXTc0oALXEw?

Part of this weeks All In sounded like territory we covered a year ago in Episode 1 of the

Friends of the pod knew this already 👀

Great chat with

for the Big Episode 50 on the

!

Nice deep dive into Justin's AI psychosis and lessons learned from building about 500 projects in the last year

The Big Episode 50! Coming soon. 👀

Honestly kind of crazy that we've wracked up 50 episodes already. We must have started this pod three times but didn't find our rhythm, usually because we hated setting up in front of a camera.

50 weeks in a row, here's to the next 50!

We had

on the

this week to give us a debrief on his OpenClaw exploits.

We also got into agent autonomy vs determinism, why Pete built Wingman a "subconscious," and whether the service-as-a-software model will beat pure SaaS.

We also side quested into Australia's property cult, the fickleness of the global economy, Australia's 28 days of oil reserves, and McKinsey's AI tool getting pwned in two hours.

Another banger! 👀

Great chat lads

nostr:nprofile1qyvhwumn8ghj7urjv4kkjatd9ec8y6tdv9kzumn9wshsz9thwden5te0wfjkccte9ejxzmt4wvhxjme0qqs9dcwyxjc9rp2qayleajjygr9pvyfe34z4h8u3q2g3gsharw508kqt6e8hx 👀

have I cooked? 👀

Just ignore that the text is off centre

Supply Drop 26

Something I’ve realised from OpenClaw is most businesses are just not going to want to yolo a fully autonomous system into their business, in pretty much the same way you don’t hire a bunch of random people and let them loose.

They already have defined roles, workflows, internal controls, approval thresholds, escalation paths, and performance measures.

I’ve been experimenting with how to take a mirrored approach by giving an agent a set of hands that she can call upon depending on the work being done.

Some hands are entirely deterministic and others introduce varying degree’s of autonomy.

Some work in the business is procedural and repeatable, so it could just run as a deterministic workflow with known steps, outputs, and clear error handling.

Other work is more analytical, so it may need an agent with an objective, but it still sits inside a governed operating model.

For instance, a Transcription Hand creates podcast transcripts. A Graph Hand turns those transcripts into entities and relationships and injects them into the agent’s context graph. A Financial Hand surfaces key metrics, anomalies, and payment states from our financials.

Underneath that, the agent builds her own knowledge graph as a working memory of the business, so she can carry context forward across work instead of relying on markdown files.

She also doesn't directly execute everything herself, she reasons about what needs to be done, maps intent to known Hands or composes multiple Hands, and if no exact Hand exists, proposes a structured workflow.

The agent then becomes the equivalent of a Chief of (her) Staff who coordinates that work, tracks progress, and reports outcomes, rather than a single super-agent improvising across the whole business.

There is a Goldilocks problem here though - too much control can kill usefulness, but too little introduces unpalatable risk for a business.

I’m experimenting here, and I’m certainly not rejecting autonomy, just keen to understand how this could be applied in a way that a business already understands how to absorb.

Now, starting businesses with agents from scratch on the other hand... 🤔

This week on

,

and I chatted about why companies shouldn't be quick to outsource AI to consultants or software vendors.

We recorded this right after a 3.5 hour workshop where about 30 non-technical people built and deployed their own web and mobile apps, so it seemed like the right time to explore this.

We’re seeing a lot of organisations trying to figure out how to create an AI strategy, but for many of them, I think this is the wrong place to start.

There's a big gap in first-hand experience with AI and this is where strategy tends to break down.

It's just really hard to create a strategy for something you don't deeply understand.

It's a bit like being asked to create a strategy for magic.

Brandon Sanderson wrote his Laws of Magic as a guide for fantasy authors. If magic has no rules and can solve any problem arbitrarily, readers feel cheated. But if magic operates within clear constraints that readers understand, then solutions feel satisfying.

AI probably looks a bit like magic to many businesses right now.

When people inside a business start experimenting with the tools themselves, rather than outsourcing AI to consultants or software vendors, the organisation makes much better decisions about where and how AI fits inside the business.

This is why developing internal capability should come before strategy.

Capability means three things.

➡️ It’s shared mental models about what AI can and can't do led by internal AI champions - people who understand the technology, adapt as the technology evolves, can answer questions based on experience, and can evangelise organically.

➡️ A safe environment where this capability can be nurtured and supported over time, alongside others facing similar challenges, where gains are methodical and compound over time, and

➡️ A way to deploy what is built directly into the organisation so that capability translates into productivity across the organisation.

That’s a flywheel for capability development. It means you own your strategy and allow it to organically form and evolve over time.

It looks like Miro, acts like Miro, but its not Miro. It's Optikon, an infinite canvas that syncs with AI agents like Wingman / Claude Code, so we can collaborate visually in real time.

Uses Nostr for auth and identity, authorship, and collaboration scoped to pubkeys.

Here's a really rough early attempt from Claude Code mapping out how it might implement a JSON pipeline to collaborate with me visually on the canvas

The big episode 42 is a banger. Thanks for being our very first repeat guest on the

Yeah, I think game interfaces like starcraft, age of empires, factorio etc are really cool for visualisation and orchestration of agents. Feels perfect for our generation growing up on RTS

Level 2: Flooded Span 🎉 Complete!

Saved 11/20 lemmings (55.0%) Time: 48.2s

Play here: https://lemmings.otherstuff.ai/

#Lemmings #nostr #gaming

Day 18 of the

advent calendar. We've now dropped eighteen games making us an accidental games studio. 😂

Today we're reviving that viral classic - Flappy Bird 👀

https://advent.otherstuff.ai/

Check out my score: 854621!

At http://hangtime.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

Check out my score: 854621!

At http://hangtime.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

I scored 2118 sats playing Snake!

Play here: https://satsnake.otherstuff.ai/

#snake #nostr #gaming

Supply Drop inbound

Just scored 860 pts in Satoshi's Garden! Sats collected: 46 • Lives left: 3 Finished in 74s Character: Andy

Play at https://satsgarden.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

I scored 1035 points playing Pong and reached level 5!

Play here: https://pong.otherstuff.ai/

#Pong #nostr #gaming

Half way point of the

advent calendar and so far we've had Minotaurs, Space Invaders, Sheep Herding and Word5.

vibed up today's game - Retired in a Field - all you have to do is avoid the cows. How hard could it be? 🤔

Check out my score: 294442!

At http://hangtime.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

I scored 2481 sats playing Snake!

Play here: https://satsnake.otherstuff.ai/

#snake #nostr #gaming

I scored 2230 sats playing Space Invaders!

Play here: https://space.otherstuff.ai/

#SpaceInvaders #nostr #gaming

Today's game on the

advent calendar tests your sheep herding abilities.

Day 9 - happy herding 🫡

https://advent.otherstuff.ai/

I imagine this is how Michael Slater felt getting out on 99. Same vibes

Check out my score: 98468!

At https://hangtime.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

Check out my score: 7000!

At https://stackem.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

Check out my score: 6000!

At https://stackem.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

Check out my score: 3300!

At https://stackem.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

Just skied 20315m in Ski Sats! Sats collected: 47 Survived for 1m 25s

Play at https://skisats.otherstuff.ai

Race the slopes in Ski Sats: endless skiing, sats collecting, and yeti dodging.

Does anyone remember playing SkiFree back in the 90's? Was aiming for a nostalgia hit by vibing up SkiSats. Go a for a festive ski, collect sats and you may or may not get chased by a yeti. 👀

https://advent.otherstuff.ai/

Just skied 3580m in Ski Sats! Sats collected: 11 Survived for 28s

Play at https://skisats.otherstuff.ai

Race the slopes in Ski Sats: endless skiing, sats collecting, and yeti dodging.

Check out my score: 79324!

At https://hangtime.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

Check out my score: 12718!

At https://hangtime.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

Just scored 820 pts in Satoshi's Garden! Sats collected: 47 • Lives left: 3 Finished in 71s Character: Andy

Play at https://satsgarden.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

For a while I wondered if I could make myself a playable character in a video game, so I vibed up a Super Mario style platformer for day 5 of the

advent calendar.

For listeners of

, this game will probably sound familiar. You can play as yours truly or Pete Winn and the goal is to collect Sats and avoid pesky central bankers that look remarkably like Janet Yellen. 👀

http://advent.otherstuff.ai

Fire up Satoshi's Garden at Day 5 of the Advent Calendar. FYI there's one trap in this game that I left in. Enjoy! 😉

Check out my score: 7190!

At https://hangtime.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

Another day another vibe coded game dropped in the

Advent Calendar 👀

A couple of big games coming in the next few days! 👀 👀

https://advent.otherstuff.ai/

Check out my score: 3339!

At https://blocks.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

Everyday in the lead up to Christmas,

and I are releasing a new game in our

Advent Calendar.

Today's game - Minotaur Labyrinth - was made by Pete and his son over breakfast

My high score is 3980 just quietly 👀

https://advent.otherstuff.ai

Check out my score: 3980!

At https://minotaur.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

Check out my score: 3741!

At https://minotaur.otherstuff.ai

Check out more games daily this december at https://advent.otherstuff.ai/

Just going to leave this here 👀

Check out my score: 1551!

At https://spider.otherstuff.ai

Check out more games daily this December at https://advent.otherstuff.ai/

We often speak about just how much more you can get done with AI agents, so in the spirit of 'show, don't tell',

and I are releasing a game a day in December to prove it.

Check out our new advent calendar of vibes:

https://advent.otherstuff.ai

SME owners aren't adopting AI, and it turns out it's the same reason they don't work on succession thinking.

They're busy.

This was good stuff

Just quietly, today's

was a banger. Out tomorrow 👀

Friday night challenge - can I clone a Framer template with Codex in under an hour?

The answer is more or less, but polishing it to my liking took a couple of hours. https://blossom.primal.net/30fd87ef0c6c0b4a1c171d28dbb28e4e07fa736f6d1b42297f08c16e175d7887.mp4

It was time for a fresh coat of paint. Coming soon 👀