Founders' Journey - Episode 125
Context is king and the evolution (not revolution) of roles
Welcome back folks to another fantastic Friday. I've just finished filming for the Insights podcast we do with FEX Global this morning, done another episode of the Digital Nexus podcast return with 🤖 Chris Sinclair and also dusted off another Redbelly Network Insights episode with Matthew Hale yesterday and an Australian DeFi Association meetup on Wednesday over at Stone & Chalk (thanks to our guests and sponsors including Zann Maxwell from City of Sydney, Ian Buswell from BLOCKv, Olga Prokhorova Matthew Brown repping Bybit and Lisa Wade (she/her) of Fableration for dialling in).
In short, it was a big week and despite all of the events above, we still had time to consult on projects, further develop our solutions like sike.ai and check out the news.
Speaking of, 2 things stood out this week including a great podcast on Context Engineering (which is very well aligned with Sike) and an article on the evolution of roles from the Financial Times.
Why Context Engineering > Prompt Engineering
In the world of LLMs, it's tempting to believe that crafting the perfect prompt is the holy grail. But a recent talk on the “12-Factor Agent” framework flips that idea on its head. The real lever for building reliable, high-performance AI agents? Context engineering.
Here's a video that inspired this:
Unlike humans, LLMs are stateless — they don’t retain memory across interactions unless you explicitly pass it in. That means the quality and structure of what you feed them (not just how cleverly you prompt them) is what determines the quality of the output.
“LLMs are pure functions. Token in, token out. The reliability of your agent comes down to the quality of the context you put in.” — 12-Factor Agent talk
Context engineering is about owning what gets passed into the model — which documents, which task instructions, which prior outputs — and doing it with precision, clarity, and relevance. It’s not just about cramming more in. It’s about sending the right information, structured in the right way, for the model to perform.
Sike Was Built with This Philosophy from Day One
At NotCentralised, this approach deeply resonated because it's exactly how we designed Sike.
We’ve always believed that giving users better control of context — not just over prompts — would unlock more reliable and trustworthy agent behaviour. Here’s how we’ve operationalised that belief:
Projects: The Organisational Layer for Context
Rather than treating documents and notes as ad hoc inputs, Sike lets you group them into Projects. Each project becomes a curated context universe — complete with documents, notes (“Dreams”), and chat threads that grow over time.
This structure is what makes context engineering usable for real people. You’re not building prompts from scratch — you’re working with well-organised, persistent context.
Recalling Context Intelligently
Sike makes it effortless to recall relevant context across your AI workflows:
Agents can pull from Documents and Dreams specific to their task.
You can define Global Variables that feed context across the entire workflow.
Conversations within a project can now reference prior threads — ensuring the AI understands ongoing discourse, not just isolated prompts.
This means your agents aren’t guessing — they’re grounded in the same thread of reasoning that you are.
Verifiability through Context Tracing
Context isn’t just passed in — it’s auditable. Our AI Check feature allows users to see exactly what parts of a document the AI relied on, and how closely they match the output. It’s context transparency — built-in.
Designed for Builders and Non-Technical Users
Whether you’re a developer chaining agents together or a domain expert running research workflows, Sike gives you fine-grained control over what the AI sees — without asking you to be a prompt engineer. You’re shaping context, not wrestling with prompt templates.
In Summary
What the 12-Factor Agent framework highlights — and what we see every day in real-world usage — is that context is the new code. The better you manage it, the more reliable and useful your agents become.
And that’s why Sike isn’t just an AI wrapper. It’s a context engine — built from the ground up to let you shape, reuse, verify and evolve what your AI knows, and what it does with it.
AI Is Just a Tool — The Jobs Will Evolve, As They Always Have
A recent piece in the Financial Times offered a sharp but reassuring perspective on the future of work in an age of AI. It reminded us that what AI changes are tasks, not entire jobs — and the outcome depends entirely on how we use the technology.
Link to FT article: https://www.ft.com/content/b67bcbaa-67b9-48f8-ad4d-4cd6f9bdc1b7
Take the example of the gardener. While it might seem like robots and sensors are coming for the shovel, the truth is more nuanced. AI may automate irrigation or detect pests — but the essence of the role evolves rather than disappears. Communicating with clients, designing spaces, applying intuition — these are still human strengths. And sometimes, ironically, it’s not the manual labour that's most draining, but writing emails or updating records — tasks where AI can genuinely help.
The article draws on historical examples too — contrasting how accounting clerks flourished thanks to spreadsheets (which removed the boring bits and elevated the thinking work), while inventory clerks were gradually deskilled by automation that removed the core expertise from their roles. The takeaway? It’s not just what gets automated — it’s what remains.
This is where our view at NotCentralised is clear:
AI is just a tool. A powerful one, yes — but like any tool, it can be misused, underused, or overhyped. It can make a bad workflow worse, or a good one more efficient. The key isn’t resisting it — it’s learning how to use it well.
We’ve been here before. From the printing press to the PC, from spreadsheets to cloud software — every new wave of technology has prompted hand-wringing about job losses. But humans adapt. We reshape our roles, refocus our energy, and expand into the spaces machines can’t fill.
We’ll do the same with AI.
The real question isn’t “Will AI take my job?” It’s “What parts of my job are worth keeping — and how do I use AI to elevate them?”
That’s why, in tools like Sike, we’re not trying to replace people. We’re building systems that give you more control, more clarity, and more leverage — so you can spend your time on the parts of the job that matter. Less admin. More insight. Less busywork. More breakthroughs.
Just like any era before this one, the future of work won’t be about man versus machine. It’ll be about how we partner with technology to push what’s possible — and find meaning in the parts it leaves to us.
Anyway folks, that's it for this week. Hope you have a great weekend and I'll see you next time.
Mark