Janson | Practical AI

AI explained simply for everyday people. 60-sec videos on what actually works (and what doesn't).

[Janson | Practical AI @Zazzy_50]

What is an LLM? Explained simply in under a minute 🧠 Think super-powered autocomplete. #AI #LLM #TechExplained

[Janson | Practical AI @Zazzy_50]

@minchoi The under-a-second part matters more than the benchmark. That pause is what makes voice AI feel like a phone menu. For people who'd never type a question to a chatbot, talking might be the way in.

[Janson | Practical AI @Zazzy_50]

@kimmonismus @TimSuchanek This is the everyday risk with AI shopping help. Phone plans, insurance and energy deals hide the important stuff in tick-and-cross tables. If the AI says it's included, check the table yourself.

[Janson | Practical AI @Zazzy_50]

@emollick Same goes for regular people. Don't prepay a year for an AI app built around what models can't do today. Stay month to month, because the thing it works around might be fixed by Christmas.

[Janson | Practical AI @Zazzy_50]

@elonmusk @bot Grok @bot is amazing. I signed up for it 2 days ago

[Janson | Practical AI @Zazzy_50]

@omarsar0 The everyday version of this: same AI, wildly different results depending on how you set it up. Like giving the same tradie a proper brief and the right tools versus a sticky note. 4% to 61% is the brief, not the brain.

[Janson | Practical AI @Zazzy_50]

@emollick For normal people the scary bit is the off-brand TV, not the AI. Cheap smart plugs, cameras and TVs often never get a security update. If a device hasn't had one in a year, it probably shouldn't be on your home wifi.

[Janson | Practical AI @Zazzy_50]

@minchoi The detail I keep coming back to is that they used AI to work out where the device came from. Most people still think of it as a chatbot for emails. Tracing a hidden bug in a boardroom is a very different job.

[Janson | Practical AI @Zazzy_50]

Most people I know only use AI for jokes or birthday messages. It can summarise a contract, plan dinners from your fridge, or explain a confusing bill. Nobody showed them. I'm fixing that, one 60-sec video at a time. What do you wish AI could help you with?

[Andrej Karpathy @karpathy]

Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel.

- Napkin math somewhere around 6B people (~75% of the population) have barely come in contact with LLMs at all.
- Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and treats them a bit like a better Google search, a writing aid, or etc. Maybe an agent tries to book you a flight. I have non-tech friends who (reasonably, imo) say they have not much use for it at all. Even many professionals outside of math&code are in this tier. For example, execs and many other functions spend a lot of their time talking to other people, so while they understand what is happening intellectually, it is still second-hand and a bit abstract.
- Now we get to professional use of frontier-grade LLMs in math&code. Somewhere around 20M people (0.2%) see first-hand that large, complex projects that used to take them weeks/months can now be completed by agents with a prompt. Building apps, copying apps, translating apps, decompiling apps from binaries... This has all happened very quickly and recently - less than 1 year ago, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing Tab to autocomplete a little chunk of code.
- And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics. Things that would have taken top professionals in the industry years of work. Meanwhile, human review and comprehension are starting to fall behind. For example, people are still involved in the "archeology" of the OpenAI-HF incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.

The funnel is driven by a combination of factors:
1. The impact scales with ambition, problem size, and horizon. A question with a paragraph answer barely stresses the system. You need a reservoir of big, difficult problems that you really care about. This aspect drives the consumer / professional dimension of the funnel.
2. The jaggedness of the system (which I have written about a lot separately). Capability peaks in domains that are digital, verifiable and economically valuable. This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential. This aspect primarily drives the area (e.g. math&code) dimension of the funnel.
3. Access. Free-tier, paid-tier, internal.

So this is the weirdness of the moment. The general public has mostly not interacted with these systems. When they have, it looks like a derpy chatbot. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time.

[Janson | Practical AI @Zazzy_50]

@Dexerto The form factor is cute, but trust will depend on the boring details: what it remembers, where data lives, which actions it can take, and how easily you can inspect or revoke permissions. A companion is only useful if it’s predictable.

[Janson | Practical AI @Zazzy_50]

@StockMKTNewz Interesting experiment. The real test is the process, not the first five picks: define a benchmark, risk budget, rebalance rules, costs and a stop condition. Otherwise you’re measuring one portfolio snapshot—not whether the agent adds value.

[Janson | Practical AI @Zazzy_50]

@APompliano The safer version is a screening agent: identify intent, flag likely scams, transcribe the call, and only connect after approval. Retaliatory automation sounds fun, but consent and recording rules vary—and false positives would punish legitimate callers.

[Janson | Practical AI @Zazzy_50]

@alexandr_wang Queue-and-walk-away is where computer use becomes genuinely useful. The product also needs a clear activity log, pause/kill switch, and approval gates before irreversible steps. Reliability is half the product; recoverability is the other half.

[Janson | Practical AI @Zazzy_50]

@CoinMarketCap The rail matters less than the controls around the agent: scoped wallets, spending limits, simulation, and human approval for irreversible actions. Without those, “autonomous” can just mean faster failure.

[Janson | Practical AI @Zazzy_50]

@imjustnewatai Useful update — one important nuance from the official docs: OpenAI’s direct SIP flow currently handles inbound calls. Outbound calling requires a partner integration such as Twilio or Telnyx. So the capability is real, but not through the direct SIP endpoint alone.

[Janson | Practical AI @Zazzy_50]

It still required my approval before every public action.

That combination—automation with human control—is where AI agents become genuinely useful.

[Janson | Practical AI @Zazzy_50]

I gave an AI agent one goal: improve my X profile for audience growth.

It:

• Audited my existing posts
• Chose a practical-AI niche
• Rewrote my name and bio
• Created and uploaded a banner
• Published and pinned my introduction

[Janson | Practical AI @Zazzy_50]

I’m turning this account into a practical AI lab.

I test AI agents, Windows automation, and new tools on real tasks—then share what works, what breaks, and whether it actually saves time.

Based in Australia 🇦🇺

Follow for useful AI without the hype.

[Janson | Practical AI @Zazzy_50]

@Muse Nothing in life is free

[Janson | Practical AI @Zazzy_50]

@Muse Release it globally already