What Claude Can't See — and What to Do About It

Something happened earlier this year that I've been thinking about ever since. I was working with claude.ai on an outreach letter — a carefully considered piece of communication with real stakes. We'd drafted it, reviewed it, refined it. Then new information arrived. A single fact I hadn't known. Within moments, the letter we'd built together looked completely different. Not because anything we'd written was wrong. Because the box had shifted. That's what I want to talk about. The box Claude doesn't have persistent memory. It doesn't browse the internet in real time. It doesn't know what's happening outside the conversation. What Claude can see, reason about, and work with is bounded by what's currently in context — the conversation, the documents, the facts you've shared, the framing you've established. That boundary is the box. This isn't a Claude quirk. It's how all large language models work. The box just has different dimensions depending on the tool. Most people understand this at some level. What's less understood is that the box is not fixed. It shifts. And it shifts through three distinct mechanisms — only one of which most people think about. Three ways the box moves New external data. Information arrives from outside that wasn't there before. You share a document, a link, a finding. The box expands. Claude can now see things it couldn't see before, and the work adjusts accordingly. This is the mechanism most people imagine when they think about context. Previously known, newly relevant. The information was already in the box. But something — a new question, a different angle, a connection that wasn't visible before — makes it suddenly significant. No new data arrived. The box reconfigured around what was already there. Human input. You tell Claude something. It might be accurate. It might be mistaken. It might be incomplete. It might be deliberately deceptive. From inside the box, all three of these look identical in the moment they arrive. The box adjusts either way. That last one is the one worth sitting with. The key insight Claude cannot distinguish a correction from a deception when it arrives. Both shift the box. Both change what Claude sees. Both change what Claude produces. That's not a malfunction — it's a structural feature of how the system works. What follows from that is important: the human in the collaboration carries the responsibility for what goes into the box. You are not just directing the work. You are curating the context. What you put in, how you frame it, what you withhold — all of it shapes what Claude can see. A skilled practitioner manages the box deliberately. An unskilled one manages it accidentally. What this means for "hallucinations" Most explanations of AI hallucinations make them sound like lying, or malfunction, or a model that simply invents things. That framing makes the problem feel unpredictable and the tool feel untrustworthy. Here's a more accurate frame: an AI error is often the result of data selection happening in an instant — inference filling a gap the model didn't know was there. The reasoning is coherent. The box was incomplete. The output reflected the box. That's a different problem than "the AI made something up." It's a problem of curation. And curation is a human skill. The practical implication: before you ask what went wrong with the output, ask what was in the box when it was produced. Often that's the answer. The box shifts inside a session too Here's something that happened to me recently that illustrates the subtler version of this problem. Early in a working session, I asked Claude to produce a document I could use in Word — a specific deliverable, clearly stated. Claude confirmed. We moved on. Over the next hour, other work accumulated in the conversation. New decisions were made. New context arrived. When the document came, it was a Markdown file. The commitment hadn't been forgotten. It had dropped out of the visible portion of the box as context accumulated. Claude wasn't being unreliable. The box had shifted, and the earlier agreement had moved out of frame. This is the version of the problem that catches serious practitioners off guard — not the obvious gaps, but the accumulated drift. Two things address it: Standing rules loaded at session open. If the commitment matters across sessions, it belongs in a document that gets loaded at the start of every session — not left to persist in a single conversation thread. Explicit confirmation before delivery. "I'm about to send you a Word file — confirm this is correct before I proceed." A two-second checkpoint before action. The cost is nearly zero. The catch rate is high. Both solutions come back to the same principle: the human manages the box. The tool works within it. Managing the box deliberately A few things that follow from this model: When you correct a fact mid-session, don't assume the correction propagated. Claude doesn't maintain a unified working model — the corrected version and the original may both exist in context. Ask explicitly for an audit of all instances. What you withhold shapes the output as much as what you include. A box that's missing a constraint produces work that ignores the constraint — not because Claude failed, but because the box didn't contain it. Clean context produces clean work. The infrastructure I've built for multi-session AI collaboration — handoff documents, decision logs, session protocols — exists largely to keep the box well-curated at every session open. The forgetting isn't the problem. A poorly curated box is the problem. The system, briefly The box model is one of the principles I've been developing while building a real publishing project with Claude as my collaborator — a series of consumer reference guides, now spanning multiple states. The architecture that makes serious multi-session AI work possible is becoming a book: Building a Book With Claude, a practitioner's guide for anyone doing long-form work with an AI tool. If you're working seriously with AI and want to understand the mechanism, not just the output — watch for it. In the meantime: you manage the box. The work will follow. Do you have standing rules for your use of AI? I'd like to hear about them. Find me on LinkedIn at linkedin.com/in/jayelkes — or connect if this problem is yours too. Jay Elkes finds the simple solution hiding inside the complicated problem. He has a talent for finding the right shoulders to stand on. He's been assembling a toolkit for doing that — across domains, in practice — long enough to give it a name: Extreme Common Sense. Follow along at extremecommonsense.net.

Raising Your Standards

Success is about staying consistent. Staying consistent is about standards. Poor standards limit success. If you raise your standards, you raise your consistency and your success--the triangle of success grows.

Darren Hardy recently posted a video with his thoughts on the value of consistency in achieving success. I think that 's only true if you can consistently raise your standards to their highest level. It is easy to be consistent at a level that doesn't achieve success.

As a programmer, my first programs were good enough to meet my client's requirements and my boss's expectations. If I had stayed consistent at it, I would have achieved a limited version of success. Many of my peers did exactly that.

I started to recognize a model within the programs I wrote. I could copy a program that was already working, delete its unique features, replace other items in the original program with their analogs for the new one, and write the unique part of the new program. An inventory update program became the start of an accounts receivable update program. This cut my writing time in half and was my standard practice for a while.

Then I realized that changing the "inventory" identifier to "account" added no value but introduced changes throughout my model. I could apply the changes consistently or eliminate the labels and make the code itself consistent. I did neither.

Instead, I eliminated the labels and pulled all the consistent code into a standardized set of modules. Then I built a new standard reference program to tie the pieces together in a standard way, and built a utility to make changes needed in the reference program. This package became my new standard, and with it a complex update program became a consistently reproducible process anybody could use. I had a new standard, a new level of consistency, a new measure of success. Work that took weeks could be dome in minutes.

Hardy's video refers to the familiar story of the tortoise and the hare. It is a great place to start. To that familiar race, let's add a puppy, who at birth can barely move and needs his mother to survive. Day by day, his consistent growth makes him bigger, stronger, faster.

Success can be achieved by being consistent, but only if the level of consistency is high enough (a good enough standard) and you have enough time. A rabbit lives twelve years, a dog  twenty, a tortoise two hundred. How much time do you have for success. How long do you want it to take?

In our competitive world, yesterday's high standard may be barely adequate today and irrelevant tomorrow. The world is littered with standards that have been made obsolete -- VHS tapes and floppy disks being obvious examples. Continuous improvement needs to be a goal we aspire to even for our standards.

I was never an athlete as a child and spent most of my adult life as a couch potato. A few years ago, I changed my habits, raised my standards, and lost 35 pounds. They stay off because my new standard is high enough to maintain my weight. If I want more, I need to raise my standard. Perhaps you do too.

Book Review: The Art of Learning

When I mentioned to a friend that I was reading the book The Art of Learning by Josh Waitzkin, she asked why. "Is there anything you don't know about learning?"

I think of myself as reasonably smart, but I used that gift to get through school. As an adult, I have come to understand the value of leveraging my strengths and therefore look for small improvements on things I do well. 

Is there anything I don't know about learning?  Since I don't know what I don't know, all I can do is assume there is. 

As Waitzkin points out:

"..:there are clear distinctions between what it takes to be decent, what it takes to be good, what it takes to be great, and what it takes to be among the best. If your goal is to be mediocre, then you have a considerable margin for error."

I was particularly intrigued by a 3 step approach to resilient, self-sufficient performance. 

1. Flow with the distraction, like a blade of grass bending to the wind. 

2. Use the distraction, inspiring ourselves with what initially would have thrown us off our games. 

3. Recreate the inspiring settings internally. 

The book covers developing mental skills, especially focus, to compete at elite and world-class levels as illustrated by the author's experiences in two different competitive arenas. 

Book Review - Deep Work by Cal Newport

Whether you read David Allen on stress-free productivity, Brian Tracy on eating the biggest frog first, or almost anything in the domain of personal productivity, the theme is juggling your actions. Newport argues that for people dealing with intense
complexity or creativity you need to carve out significant blocks of time for what he calls deep work.

This isn't about multitasking or task switching. The book focuses on why and how to create an environment where deep work is possible, and shares stories of people who have done it successfully. He gets into more detail, but the key ideas are block out the time and cut out the distractions.

You won't find anything new or astonishing on blocking out time here. Newport methodically walks through all the logical possibilities but in the end it is a matter of what works.

As for cutting out distraction, there is a lot of actionable advice, some of which may call for serious reflection. High on that list is cutting out social media. You may not be ready to do that, but perhaps you could get them off your phone and out the way. Newport walks his talk here--no Twitter account.

Not everybody needs to do deep work, but there are fewer exceptions than we realize. The book didn't help me block out time, but it did help me think.


- Posted using BlogPress from my iPhone

False Choices

"Better 100 friends than 100 rubles."~ Russian proverb

I don't always agree with proverbs, and I don't agree with this one. While I believe that friends are more important than money, I think the proverb is inspiring a false choice. Both are good and you can have both.

In his book "Secrets of the Millionaire Mind" T. Harv Eker points out that rich people think "and", the rest of us think "or". By thinking or, that we must choose one and not the other, we are making a choice when no choice is required--a false choice.

We can also make a bad decision by using or to justify a bad decision. Should we eat the fries or the dessert? Neither. Sometimes the right choice is to do nothing, or at least none of the above. When at a restaurant, look for what you'll wish you had eaten, not for what looks good. It all looks good, right.

As a computer programer, I learned that "and", "or" and not are logical filters that can precisely control what to include or exclude. When they are needed and used appropriately good things happen.

So listen for the words: and, or, not. Ask yourself why, and why not.

Fries or dessert? Why not neither?

100 friends or 100 rubles? Why not both? Why not 1000?



- Posted using BlogPress from my iPhone

The School of Greatness

For the past few years, Lewis Howes has been hosting a podcast he calls the School of Greatness. During a typical program he will interview an expert on personal development or some related topic. His skill as a podcaster, the quality of the guests, and the value to listeners continue to go through a positive feedback loop.

Many of his guests are authors, trainers and educators--what else would you expect at a school? His goal us to help his listeners develop a lifestyle of their win choosing. Now he has released a book with the same goal. For example:

When you want to lose weight and keep it off, you don't go on a diet, because diets are about artificial restriction. They're miserable. Instead, change your lifestyle to match your goals.

This book is about changing your lifestyle to match your goals, a step by step process that can help you take control of your life and destiny.

The School of Greatness: A Real-World Guide to Living Bigger, Loving Deeper, and Leaving a Legacy https://www.amazon.com/dp/1623365961/ref=cm_sw_r_awd_VWsowb0XQAQDD

Why I Subscribe to Podcasts

As far as I see it, there are two types of people: those who think they can multitask and those who know better. With minor exceptions, multitasking breaks focus and slows productivity. This is one of those exceptions. 

You can call it a habit, a goal, a standard or crazy, but I try to walk at least 10,000 steps--about five miles--every day. In the past month I have done this 28 of 31 days. As I walk I listen to podcasts. I subscribe to several and they fall into three categories. 

News 

About half the walk is an audio version of the PBS News Hour. Once a week, I satisfy my inner science geek with This Week in Science. These and a couple others remind me that there is an

outside world. 

Personal Growth

I pick up an assortment of personal

growth podcasts doing interviews or offering ideas. Some of these lead to books to read, either in full or through a Blinkist summary. If I'm not hooked in the first couple minutes, I'm on to the next topic. 

Other 

This category covers hobbies and entertainment. 

Why

By playing these podcasts while I walk, I can get a briefing on current events, pick up some useful ideas and catch a few messages about my hobbies while allocating the time to health and exercise. I generally play the podcasts at 1.5 speed, as fast as possible without distorting voices. 

Truthfully, this time is for the exercise. The podcasts are scanned for potential interests but not rigorously studied. Over the past few years that I've done this, the News Hour has dutifully reported the Dow Jones Industrial Average over 1000 times and I couldn't guess it within 10% or even define it accurately. I guess I was right--I can't multitask, but it does make walking those miles more pleasant. 

Your Turn

What podcasts do you follow and why? Are you giving them your full attention?