Category: What Dimitri Learned

I’ve been spending most of my hobby time playing with reasoning boxes and it’s been crazy fun. I need to write properly about my explorations later, but so far, I’ve convinced an LLM to brainstorm ideas and then critique itself in a diverge-converge exercise, evaluate Cynefin space of a situation, and even get some Humean reasoning going – albeit with mixed results on the last one.

One pattern that I’ve leaned on consistently is asking an LLM to produce output in JSON. There are several advantages to that. First, I get the output that I can then process outside of an LLM. Second, JSON surprisingly acts as a useful set of reasoning rails for LLMs itself. It kind of makes sense – since LLMs are predictive devices, laying out a structure of prediction helps organize and guide the prediction itself. In other words, JSON is not just useful for processing the output. It also helps with framing the process of text completion.

Of course, now that we’re asking an LLM to return JSON, the question arises: how do we inform it of the structure in which this JSON needs to be?

The first obvious answer is that we already have a way to describe the structure of JSON output. It’s called JSON Schema and it’s in broad use across the industry. Turns out, modern LLMs will happily consume JSON Schema. All you have to do is, somewhere in your prompt, tell them something along the lines of (you can see an example here):

Respond in valid JSON that matches the following JSON schema:
<insert your JSON schema here>

The output will (well, with 95% probability, as it usually goes with LLMs) dutifully follow the specified schema.

Unfortunately for us, JSON Schema is a bit chunky. Designed for precision and processing, it’s verbose and can quickly blow through our token budget if we’re not careful. It is also not exactly a human-readable format, making prompt hacking a bit of a chore.

On the other hand, describing JSON in English is also a bit too weird: “the output must be valid JSON containing an object with foo and bar as strings, blah blah blah” is something that LLMs will tolerate, but who has the time for typing it all in prose?

So I ended up using this weird pidgin that I named Schemish. It’s not really a JSON Schema, but not plain English either. Instead, it’s a mock of JSON output, with hints for the output provided as values. Kind of like this:

Respond in valid JSON of the following structure:
{
  "context": "A few sentences describing the context in which the question was asked",
  "subjects": [
    {
      "name": "Name of an subject mentioned or implied in the question",
      "assumption": "An assumption that you made when identifying this subject",
      "critique": "How might this assumption be wrong?",
      "question": "A question that could be asked to test the assumption",
    }
  ],
  "objects": [
    {
      "name": "Name of an object mentioned or implied in the question",
      "assumption": "An assumption that you made when discerning this object",
      "critique": "How might this assumption be wrong?",
      "question": "A question that could be asked to test the assumption",
    }
  ],
}

As you can see, I am specifying the structure of the output, but instead of relying on JSON schema syntax, I am simply mocking it out – just as I would describe JSON output in a sketch for another person.

If we want to go even more compact, we could go for YAML:

Respond in valid YAML of the following structure:
---
context: "A few sentences describing the context in which the question was asked",
  subjects:
    - name: "Name of an subject mentioned or implied in the question"
      assumption: "An assumption that you made when identifying this subject"
      critique: "How might this assumption be wrong?"
      question": "A question that could be asked to test the assumption"
  objects: 
    - name: "Name of an object mentioned or implied in the question"
      assumption: "An assumption that you made when discerning this object"
      critique: "How might this assumption be wrong?"
      question": "A question that could be asked to test the assumption"

Now, for an even cooler trick. You might have noticed that the structure has somewhat repeating elements in it. JSON Schema has references and other tools to deal with redundant declarations. Turns out, Schemish can do the same thing. Just use what you would usually do when sketching out a structure for yourself or your colleagues – leave a hand-wavy comment:

Respond in valid YAML of the following structure:
---
context: "A few sentences describing the context in which the question was asked",
  subjects:
    - name: "Name of an subject mentioned or implied in the question"
      assumption: "An assumption that you made when identifying this subject"
      critique: "How might this assumption be wrong?"
      question": "A question that could be asked to test the assumption"
  objects: 
    # same structure as `subjects`

What’s neat about Schemish is that it’s clearly derivable from a JSON Schema. Though I haven’t actually written the code, it’s fairly obvious that JSON Schema description fields as well as type and optionality can be expressed as English (or some shorthand of it) in the value of the Schemish structure.

This means that I could potentially connect a piece of machinery that speaks JSON Schema to an LLM via Schemish, with Schemish translator shepherding the schema from precise, mechanical code-land to the squishy, funky LLM-land – and then picking up the output, produced by the LLM to usher it back to the code-land by validating it with JSON Schema.

This Schemish idea seems fairly useful and I’ve seen similar techniques pop up in the AI tooling sphere. If you like it, please do something with it. I’ll probably write a simple wrapper for Schemish, too.

This one is a very short framing, but hopefully, it’s still useful. Whenever we operate in environments that look like pace layers (and when don’t they?), there emerges a question of relative velocity. How fast am I moving relative to the pace layer where I want to play?

If the environment that surrounds us is moving slower than our team’s velocity, we will feel overly constrained and frustrated. Doing anything appears to be laden with unnecessary friction. On the other hand, if the environment is moving faster than us, we’ll feel like a tractor on an autobahn: everyone is zooming past us. Whenever something like this happens, this means that our relative velocity doesn’t match the velocity of the pace layer in which we’ve chosen to play.

This mismatch will feel uncomfortable, and there is a force hiding behind this discomfort. This force will constantly nudge us toward the appropriate layer.

If, after working in a startup where writing hacky code and getting it out to customers quickly was existential, I join a well-structured team with robust engineering practices and reliability guarantees, I will feel the force that constantly pushes me out of this team – the “serious business” processes will aggravate the crap out of me.

Conversely, if I grew up in a large company where landing one CL a week was somewhat of an achievement, joining a small team of rapid prototypers may leave me in bewilderment: how dare these people throw code at the repo without any code reviews?! And why does it feel like everyone is zooming past me?

Similarly, every technology organization has a waterline. Everything above this waterline will move at a faster pace than the organization. Everything below will be slower.

If this organization decides to enter the game at the faster-pace layer, they will see themselves being constantly lapped: new, better products will emerge faster than the team can keep up with. We would be still in the planning stages of a new feature while another player at this layer already shipped several iterations of it.

Similarly, the organization will struggle applying itself effectively at the lower pace layers. We would ship a bunch of cool things, and see almost no uptick or change in adoption or usage patterns. Worse yet, our customers will complain of churn and instability, choosing more stable and consistent peers over us.

No matter how much we will want to play above or below our waterline, we’ll keep battling the force that nudges us toward us. We might speak passionately about needing to move faster and innovate, and be obsessed with speed as a concept – but if our team is designed to move at a deliberate pace of a tractor, all it will be is just talk.

When an organization is aware of its waterline, it can be very effective. Choosing to play only at the pace layer where the difference in relative velocity is minimal can be a source of a significant advantage in itself: the organization can apply itself fully to solving the problem at hand, rather than battling the invisible force that nudges it out of the pace layer it doesn’t belong to.

I’ve been thinking about the boundaries of what’s possible when applying large language models (LLMs) to various domains. Here’s a framing that might not be fully cooked, but probably worth sharing. Help me make it better.

To set things up, I will use the limits lens from the problem understanding framework. Roughly, I identify three key limits that bound what’s possible for anyone who is put in front of a problem: capacity, or the actual ability to find effective solutions (one or more) to the problem; time, the velocity at which effective solutions are discovered; and finally, attachment, or the resistance to incorporate interesting information into their understanding of the problem to find solutions.

These three limits govern what we humans can do. Let’s see how these limits could be applied to LLMs.

Let me start with capacity. The fact that this limit applies to LLMs is fairly self-evident, and can be illustrated easily with the progress we have made in the last year. What was impossible just 18 months ago is now a widely accepted capability. The progression to larger models, larger context windows feels like a constant drumbeat – this is the capability limit being pushed further and further back. When will we run into the wall of the invisible asymptote? When will we be able to confidently say: “Aha! Here’s proof that LLMs can’t actually do that”? It’s not easy to tell. However, just like with all things of this world, this limit is still present with LLMs. We just haven’t quite found it yet.

Another interesting trick that LLMs have compared to humans is the ability to clone. A bot or an agent or a network of reasoning are easily reproducible. One does not need to spend a lifetime raising an LLM from infant to adulthood. Once a pattern is established, clones of this pattern are easy to produce. This is a significant source of capacity. Being more numerous is easy, which means that brute force can compensate for smarts in a pinch.

The final component of the limit of capacity is the computing power. The amount of energy it takes to make the LLM produce a potential answer to a given problem seems like an important factor. Again, we seem to be marching along the Moore’s Law curve here, and I expect each new breakthrough in atoms and bits to significantly push the capacity limit out.

Speaking of the future, let’s talk about the limit of time. I sprinkled the references to time into my examination of the limit of capacity. They seem closely interrelated. LLMs becoming more capable and efficient means that they will also be able to solve problems more effectively. At this moment of time, I am contemplating networks of reasoning boxes, which implies the era when invoking LLMs will take single milliseconds. This seems to fall out naturally of the limit of capacity advancing. There will likely be an asymptote we’ll hit with how fast an LLM can go, but we’re definitely not there yet.

The limit of attachment is the one that’s been most curious for me. As far as I can tell, LLMs don’t seem to have it at all. While people could get bored or tired, or have anxiety about doing things one way or another, express strong convictions and make rash decisions… LLMs don’t seem to have any of that. LLMs don’t have the “ooh, I wonder what <character> will do next on <TV show> episode“ mindworm. They don’t need to rush home to put their kid to bed. They have no desire to spend the day just hanging out with friends. LLMs are unattached. There is no inner shame that they have to protect at all cost. There are no values that they passionately embrace. The limit of attachment for LLMs appears to be a bottomless abyss – or a blue sky, if you’re into more positive metaphors.

Will the limit of attachment manifest itself for LLMs? Probably. There’s something very significant about this lack of the limit that indicates the vastness of the space we’re in.

So, what does this mean in practice? 

As I was saying to my colleagues, we’re at the 8-track tape stage of the whole LLM story. What we think is cool and amazing now will be viewed as silly and as quaint as the 8-track tapes just a few months in the future. Prepare for more tectonic shifts as the previously understood lines of the capacity limit are redrawn. Avoid early firm bets on the final shape of things in this space.

Also, we’re likely very close to the moment when we will have a decent representation of an “information worker”, probably as some sort of reasoning network. This worker will never get tired, will never complain about the problem being too pedestrian for their talents, will never slack off or quiet-quit. This worker will continuously improve and get better at the tasks that they are given, performing them more efficiently each time.

Looking at the LLM bounding box – or more accurately, the lack thereof – my intuition is that we’re going to see terraforming of entire industries, especially where information plumbing is a significant cost of doing business. I have no idea how they will change and what they will look like after the dust settles, but it’s very likely that this change will happen.

More importantly, as this change grinds into our established understanding of how things are, we are likely to grapple with it as a society. It is very possible that it is our own limits of attachment that we will try to impose on the LLMs, no matter how fruitless this effort will end up being in the long term.

Here’s a concrete example of a reasoning box that I’ve been talking about last week. It’s not super-flashy or cool – certainly not one of those viral demos, but it’s a useful example of recursively ground-truthing a reasoning box onto itself. The source code is here if you want to run it yourself.

The initial idea I wanted to play with was concept extraction: taking a few text passages and then turning them into a graph of interesting concepts that are present in the text. The vertices of the graph would be the concepts and the edges will represent the logical connections between them.

I started with a fairly simple prompt:

Analyze the text and identify all core concepts defined within the text and connections between them.

Represent the mental concept and connections between them as JSON in the following format:

{
"concept name": {
"definition": "brief definition of the concept",
"connections": [ a list of of concept names that connect with this concept ]
}
}

TEXT TO ANALYZE:
${input}

This is mostly a typical reasoning box – take in some framing, context, and a problem statement (“identify all core concepts”), and produce a structured output that reflects the reasoning. In this particular case, I am not asking for the chain of reasoning, but rather for a network of reasoning.

The initial output was nice, but clearly incomplete. So I thought – hey, what if I feed the output back into the LLM, but with a different prompt. In this prompt, I would ask it to refine the list of concepts:

Analyze the text and identify all core concepts defined within the text and connections between them.

Represent the mental concept and connections between them as JSON in the following format:

{
"concept name": {
"definition": "brief definition of the concept",
"connections": [ a list of of concept names that connect with this concept ]
}
}

TEXT TO ANALYZE:
${input}

RESPONSE:
${concepts}

Identify all additional concepts from the provided text that are not yet in the JSON response and incorporate them into the JSON response. Add only concepts that are directly mentioned in the text. Remove concepts that were not mentioned in the text.

Reply with the updated JSON response.

RESPONSE:

Notice what is happening here. I am not only asking the reasoning box to identify the concepts. I am also providing the outcome of its previous reasoning and asking to assess the quality of this reasoning.

Turns out, this is enough to spur a ground truth response in the reasoning box: when I run it recursively, the list of concepts grows and concept definitions get refined, while connections shift around to better represent the graph. I might start with five or six concepts in the first run, and then expand into a dozen or more. Each successive run improves the state of the concept graph.

This is somewhat different from the common agent pattern in reasoning boxes, where the outcomes of agent’s actions serve as the ground truth. Instead, the ground truth is static – it’s the original text passages, and it is the previous response that is reasoned about.  Think of it as the reasoning box making guesses against some ground truth that needs to be puzzled out and then repeatedly evaluating these guesses. Each new guess is based on the previous guess – it’s a path-dependent reasoning.

Eventually, when the graph settles down and no more changes are introduced, the reasoning box produces a fairly well-reasoned representation of a text passage as a graph of concepts.

We could maybe incorporate it into our writing process, and use it to see if the concepts in our writing connect in the way we desire, and if the definitions of these concepts are discernible. Because the reasoning box has no additional context, what we’ll see in the concept graph can serve as a good way to gauge if our writing will make sense to others.

We could maybe create multiple graphs for multiple passages and see if similar concepts emerge – and how they might connect. Or maybe use it to spot text written in a way that is not coherent and the concepts are either not well-formed or too thinly connected.

Or we could just marvel at the fact that just a few lines of code and two prompts give us something that was largely inaccessible just a year ago. Dandelions FTW.

I’ve been puzzling over a problem that many engineering teams face and came up with this metaphor. It’s situated in the general space of attachment and could probably apply to things other than engineering teams. 

Here’s the setup. Imagine that we’re leading a team whose objective is to rapidly explore some newly opened space. Everyone gets their little area of the space, and armed with enthusiasm and skill, the teams venture off into the unknown. A few months later, a weird thing happens: now we have N teams that build technology or products in basically the same space where they started.

Instead of exploring, the team just settled into the first interesting thing they found. Exploration collapsed into optimizing for the newly found value niche.

This might not necessarily be a bad thing. If the new space is ripe with opportunities or the team is incredibly lucky, they might have struck gold on the first try.  Except my experience tells me that most of the time, the full value of the niche is grossly overestimated, and the teams end up organizing themselves into settlers of a tiny “meh” value space.

The events that follow are fairly predictable. There is a struggle between us and the individual team leads to “align”, where the word “align” really stands for “what the heck are y’all doing?! we were supposed to be exploring!!” from us and “stop distracting us with your silly ideas! we have customers to serve and things to ship!” from the sub-teams. The team becomes stuck.

I have seen various ways in which the resolution plays out. There’s one with the uneasy compromise, where the “exploration team” kayfabe is played out at the higher levels (mostly in slide decks), and the sub-teams are just left to do their thing. There’s one with the leader making a “quake”: a swift reorg that leaves the sub-team leads without a path forward. There’s one where a new stealth sub-team is started to actually explore (you can guess what happens next).

The lens that really helps here is “something will get optimized”. When we have engineers, we have people whose literal job description includes organizing code into something that lasts. Like a car with unbalanced wheels, by default, engineers will veer toward elephant-land. Given no other optimization criteria, what will get optimized is the quality of the code base and the robustness of the technical solution that it offers.

The problem is, when exploring, we don’t need any of that. We need messy, crappy code that somewhat works to get a good sense of whether there’s a there there. And then we need to throw that code out or leave it as-is and move on to the next part of the space.

This is not at all natural and intuitive for engineering teams. There are no tests! Not even a code review process! This dependency was made by a random person in Nebraska! Madness!

By the way, the opposite of this phenomenon is also true. If our engineering team does not have this tendency toward building code that lasts, we probably don’t have an engineering team. We might have some coders or programmers, but no engineering.

To shift an engineering team to be more amenable to exploration, we need to shift the target of the optimization.

That’s where the cargo-and-train metaphor comes in. Let’s pretend that an engineering team is the train that delivers a cargo: a thing that it makes. The thing about cargo is that once it is delivered, it leaves the train and the train gets new cargo. Train is permanent. Cargo is transient.

To make our train go efficiently, we optimize for moving cargo as quickly as possible, and we optimize for keeping the train in its best condition. Figuring out which part of our work we optimize to keep and which one we optimize to move is what it’s all about.

If we follow this metaphor, there are two questions that an engineering team needs to ask itself: “What is our cargo?” and “What is our train?” We need to consciously separate our cargo from our train.

Which part of our business do we optimize to let go of as efficiently as possible, and which part of it do we keep and grow?

For a typical engineering team, the cargo is the software release and the code base is the train.

Each release is the snapshot of the code base at a certain state. Once that release cut is made, we mentally let go of it and start on the next release. Releasing well means being able to make a release cut like a Swiss train: always on time, with no hiccups.

The codebase is the train, since this is where releases come from. Codebase is the place where the product grows and matures. Our codebase is what we keep and improve and strive to make better with time. Terms like technical debt we engineers invent reflect our anxiety about succeeding at this process.

When the engineering team is asked to explore a new space, the answers to the two questions are like different.

It might very well be that the code we write is cargo. It’s just something we do as a byproduct of our exploration. We write a ton of prototypes, throw them in the wild, and see which ones stick.

What is the train then? My intuition is that it’s knowledge. After all, the whole point of exploration is mapping the unknown. If our continuous delivery of cargo – writing of prototypes – doesn’t light up more and more territory, we’re doing something wrong.

So when an engineering team is asked to explore a new space, we need to contemplate the cargo-and-train questions carefully and decide on our answers to them.

Then, we need to invest into making sure that everyone on the team optimizes for the right thing: the thing that we want to be our cargo is optimized to be delivered and let go of quickly, and the thing we want to be our train is carefully and lovingly grown and enriched with each delivery.

This includes everything from mission and vision where the cargo-and-train questions are clearly answered, but also into culture, incentives, and structure of the team. Remember – most of the default engineering practices and processes were designed for default engineering teams. Which means that if we’re setting out to explore, they will be working against us.

This story begins with the introduction of metacognition to large language models (LLMs). In the LLM days of yore (like a few months ago), we just saw them as things we could ask questions and get answers back. It was exciting. People wrote think pieces about the future of AI and all that jazz.

But then a few extra-curious folks (this is the paper that opened my eyes) realized that you could do something slightly different: instead of asking for an answer, we could ask for the reasoning that might lead to the answer.

Instead of “where do I buy comfortable shoes my size?”, we could inquire: “hey, I am going to give you a question, but don’t answer it. Instead, tell me how you would reason about arriving at the answer. Oh, and give me the list of steps that would lead to finding the answer. Here’s the question: where do I buy comfortable shoes my size?”

Do you sense the shift? It’s like an instant leveling up, the reshaping of the landscape. Instead of remaining hidden in the nethers of the model, the reasoning about the question is now out in the open. We can look at this reasoning and do what we would do with any reasoning that’s legible to us: examine it for inconsistencies and decide for ourselves if this reasoning and the steps supplied will indeed lead us toward the answer. Such legibility of reasoning is a powerful thing.

With reasoning becoming observable, we iterate to constrain and shape it. We could tell the LLM to only use specific actions of our choice as steps in the reasoning. We could also specify particular means of reasoning to use, like taking multiple perspectives or providing a collection of lenses to rely on.

To kick it up another notch, we could ask an LLM to reason about its own reasoning. We could ask it “Alright, you came up with these steps to answer this question. What do you think? Will these work? What’s missing?” As long as we request to provide the reasoning back, we are still in the metacognitive territory.

We could also give it the outcomes of some of the actions it suggested as part of the original reasoning and ask it to reason about these outcomes. We could specify that we tried one of the steps and it didn’t work. Or maybe that it worked, but made it impossible for us to go to the next step – and ask it to reason about that.

From the question-answering box, we’ve upleveled to the reasoning box.

All reasoning boxes I’ve noticed appear to have this common structure. A reasoning box has three inputs: context, problem, and framing. The output is the actual reasoning. 

The context is the important information that we believe the box needs to have to reason. It could be the list of the tools we would like it to use for reasoning, the log of prior attempts at reasoning (aka memory), information produced by these previous attempts at reasoning, or any other significant stuff that helps the reasoning process.

The problem is the actual question or statement that we would like our box to reason about. It could be something like the shoe-shopper question above, or anything else we would want to reason about, from code to philosophical dilemmas.

The final input is the framing. The reasoning box needs rails on which to reason, and the framing provides these rails. This is currently the domain of prompt engineering, where we discern resonant cues in the massive epistemological tangle that is LLM that give to the reasoning box the perspective we’re looking for. It usually goes like “You are a friendly bot that …” or “Your task is to…”. Framing is sort of like a mind-seed for the reasoning box, defining the kind of reasoning output it will provide.

Given that most of the time we would want to examine the reasoning in some organized way, the framing usually also constrains the output to be easily parsed, be it a simple list, CSV, or JSON.

A reasoning box is certainly a neat device. But by itself, it’s just a fun little project. What makes reasoning boxes useful is connecting them to ground truth. Once we connect a reasoning box to a ground truth, we get the real sparkles. Ground truth gives us a way to build a feedback loop.

What is this ground truth? Well, it’s anything that can inform the reasoning box about the outcomes of its reasoning. For example, in our shoe example, a ground truth could be us informing the box of the successes or failures of actions the reasoning box supplied as part of its reasoning.

If we look at it as a device, a ground truth takes one input and produces one output. The input is the reasoning and the output is the outcomes of applying this reasoning. I am very careful not to call ground truth “the ground truth”, because what truths are significant may vary depending on the kinds of reasoning we seek.

For example, and as I implied earlier, a reasoning box itself is a perfectly acceptable ground truthing device. In other words, we could connect two reasoning boxes together, feeding one’s output into another’s context – and see what happens. That’s the basics of the structure behind AutoGPT.

Connecting a reasoning box to a real-life ground truth is what most AI Agents are. They are reasoning boxes whose reasoning is used by a ground truthing device to take actions, like searching the web or querying data sources – and then feeding the outcomes of these actions back into the reasoning boxes. The ground truth connection is what gives reasoning boxes agency.

And I wonder if there’s more to this story?

My intuition is that that the reasoning box and a ground truthing device are the two kinds of blocks we need to build what I call “socratic machines”: networks of reasoning boxes and ground truthing devices that are capable of independently producing self-consistent reasoning. That is, we can now build machines that can observe things around them, hypothesize, and despite all of the hallucinations that they may occasionally incur, arrive at well-reasoned conclusions about them.

The quality of these conclusions will depend very much on the type of ground truthing these machines have and the kind of framing they are equipped with. My guess is that socratic machines might even be able to detect ground truthing inconsistencies by reasoning about them, kind of like how our own minds are able to create the illusion of clear vision despite only receiving a bunch of semi-random blobs that our visual organs supply. And similarly, they might be able to discern, repair and enrich insufficient framings, similar to how our minds undergo vertical development.

This all sounds outlandish even to me, and I can already spot some asymptotes that this whole mess may bump into. However, it is already pretty clear that we are moving past the age of chatbots and into the age of reasoning boxes. Who knows, maybe the age of socratic machines is next to come? 

So we decided to innovate. Great! Where do we begin? How do we structure our innovation portfolio? There are so many possibilities! AI is definitely hot right now. But so are advances in green technology – maybe that’s our ticket? I heard there’s stuff happening with biotech, too. And I bet there are some face-melting breakthroughs in metallurgy…

With so much happening everywhere all at once, it could be challenging to orient ourselves and innovate intentionally – or at least with enough intention to convince ourselves that we’re not placing random bets. A better question: what spaces do we not invest into when innovating?

Here’s a super-simple framing that I’d found useful in choosing the space to innovate. It looks like a three-step process.

First, we need to know what our embodied strategy is. We need to understand what our capabilities are and where they will be taking us by default.

This is important, because some innovation may just happen as a result of us letting our embodied strategy play out. If we are an organization whose embodied strategy is strongly oriented toward writing efficient C++ code, then we are very likely to keep seeing amazing bits of innovation pop out in that particular space. We will likely lead some neat C++ standards initiatives and invent new cool ways to squeeze a few more drops of performance out of the code we write.

As I mentioned before, embodied strategy is usually not the same as stated strategies. I know very few teams who are brutally honest with themselves about what they are about. There’s usually plenty of daylight between what the organization states about where they’re going and where they are actually going. The challenge of step 1 is to pierce the veil of the stated strategy.

As you may remember from my previous essays, this understanding will also include knowing our strategy aperture. How broad is our organization’s cone of embodied strategy?

At the end of the first step, we already have some insight on the question above. Spaces well outside of our cone of embodied strategy are not reachable for us. They are the first to put into the discards pile. If we are an organization whose strengths are firmly in software engineering, attempting to innovate in hardware is mostly like throwing money away – unless of course we first grow our hardware engineering competency.

The second step is to understand our innovation frontier. The innovation frontier is a thin layer around our cone of embodied strategy. Innovation ideas at the outer edge of this frontier are the ones we’ve just discarded as unreachable. Ideas at the inner edge of the frontier are obviously going to happen anyway: they are part of the team’s embodied strategy.

It is the ideas within this frontier that are worth paying closer attention to. They are the “likely-to-miss” opportunities. Because they are still on the fringe of the embodied strategy, the organization is capable of realizing them, but is unlikely to do so – they are on the fringe, after all.

It is these opportunities that are likely going to sting a lot for a team when missed. They are the ones that were clearly within reach, but were ignored because of the pressing fires and general everyday minutiae of running core business. They are the ones that will disrupt the business as usual, because – when they are big enough – they will definitely reshape the future opportunities for the organization.

The innovation frontier is likely razor-thin for well-optimized and specialized organizations. The more narrow our strategy aperture, the less likely we will be to shift a bit to explore curious objects just outside of our main field of view.

In such cases, the best thing the leader of such an organization can do is to invest seriously into expanding their innovation frontier. Intentionally create spaces where thinking can happen at a slower pace, where wilder ideas can be prototyped and shared in a more dandelion environment. Be intentional about keeping the scope roughly within the innovation frontier, but add some fuzziness and slack to where these boundaries are.

The third step is to rearrange the old 70/20/10 formula and balance our innovation portfolio according to what we’ve learned above:

• Put 70% into the ideas within the innovation frontier and the efforts to expand our innovation frontier.
• Put 20% into the ideas that are within the strategy aperture.
• Just in case we’re wrong about our understanding of our embodied strategy, put 10% into the ideas that are at the outer edge of the innovation frontier.

And who knows, my law of tightening strategy aperture could be proven wrong? Perhaps if an organization is intentional enough about expanding its innovation frontier, it could regain its ability to see and realize the opportunities that would have been previously unattainable?

Wait, did we forgo the whole notion of timelines in our innovation portfolio calculations? It’s still there, since the cone of embodied strategy does extend in time. It’s just not as significant as it was in the old formula. Why? That’s a whole different story and luckily, my friend Alex wrote this story down just a few days ago.