TL;DR

We bunched chard, onions, basil, and garlic. Photos today had a technical problem, but the old farm dog Jenkins did not. He was as attentive and ready to deliver as ever.

No Photos Today

My primary phone has suffered it’s fair share of damage over the past two years and this month has not been any kinder than the last. My prior phone’s cracked inside screen still allowed it to be usable as a ‘farm phone’ for photos and syncing up to my self-hosted Immich instance. Today I decided to use the latter of the two and none of the photos (I thought) I took showed up. A fresh new unlocked phone is on the way, but for today this means that there are no new pictures to share, despite all the recent work on the gallery to make it easily searchable by tags.

Fortunately, most of the things I’d otherwise take snaps of today were the same things we’d already been doing yesterday and last week…harvesting garlic and early onions, washing swiss chard, and bunching basil.

Jenkins is a (Very) Good Dog

Jenkins is the older of the two farm dogs. While the younger dog is about 4 years old, Jenkins is 7+ and a very good boy. He’s the one who will actively respond first to oddball sounds coming from across the field where the hens live, and though both are great little farm companions, there’s something about Jenkins that I connect with more than the younger one.

When I was a boy, we adopted an Alaskan husky pup. This was way too much dog for 7 year old me to walk and we would often find her stuck somewhere after breaking loose of whatever she was connected to when we were away. My wife grew up with border collies, the very opposite of my early dog experiences but despite that, we both share a love of animals, pets or other humankind.

Today when we were headed to lunch, I took a chance and whistled like Jamie their owner does when he wants both dogs to come up into the truck. They came immediately and we headed down, the whole (human and dog) farm crew together. I don’t like to head-fake or pretend anything and I don’t have little treats in my pocket to hand out to dogs or colleagues alike. So when I finally catch on to the behavior of creatures around me, dogs or people, to get them to understand it’s time to move on to the next thing, I feel like I’ve finally earned the right to call on them when they’re needed.

If I was a farm dog, I’d like to think I’d be like Jenkins. He is attentive, anticipates human movements, checks for human confirmation, takes initiative when appropriate, and is occasionally ornery like the best of us are. I would say this is due to the kind and patient nature of his owner alone, but I know that all dogs are different just like the farms some of them are raised on, and much of it is due to the individual dog’s personality more in the long run than their early and often short-term training.

Where Was the Training in DevOps Pipelines?

In my prior career, I spent a lot of time writing automation and helping others do so. While ‘Infrastructure as Code’ (IaC) was an important step in the right direction for its hayday, in retrospect it was never enough. The learning, the training, was only potentially in two places: 1) the human brains of the engineers who wrote the code, and 2) the proprietary platforms that had full access to anything done on them (no matter what the contracts might have said).

Most, if not all, software delivery pipeline platforms and technologies missed the mark when it came to what mattered most…even the early adopters of ‘AI’ in 2021. The pipelines we wrote were often brittle, hard to maintain, and required a lot of hand-holding to keep them running. This doesn’t work in today’s AI-infused world because much of the tuning now has to do with the prompts, the context, and the (necessarily) flexible processing patterns required to make them work reliably. The leading platforms still miss the mark by providing AI features designed to wow executive audiences, not by providing the practical features and as importantly the training needed to make them work in practice

Unlike Jenkins the farm dog, IaC-forward technologies and platforms like Jenkins, Harness, Pulumi, and certainly tech like Terraform, Ansible, and CloudFormation, are not at all self-training. They require constant human attention and effort to keep them working as expected when even the slightest things change in enterprise IT environments. One missing character from a tag deep down in a Helm chart can not only cause software delivery pipelines to grind to a halt, but possibly the production environment as well.

All I know is that if there was a CI/CD version of the farm dog, it would be attentive, anticipatory, factors in human behavior, knows when to need confirmation vs. take initiative, and only when really appropriate be ornery to us.

What My AI Doesn’t Learn Won’t Hurt It

Right now, most of the ‘learning’ in my AI pipeline is done by me, the human, and LLMs are just tools to get it to do ‘expert’ tasks faster than I can. I am thinking of doing a spike on my own embeddings this weekend and with most stuff that I think is over my head at first, will likely prove worth my efforts once I get a prototype working.

I have to do this as a next step because using other people’s stock LLMs just doesn’t get enough of the detail about organic agriculture correct. Call it a lack of training data or purposeful use cases or intentional subterfuge but not knowing a chicken coop when it’s clearly got chickens leaking out of it and is a wooden shed says a lot about what an AI-powered dog(e) was and was not subjected to in its learning phases. The outcomes of my use of generic LLMs are on this and prior posts for all to read.

Case in point, today’s AI summary way over-rotated on a last-minute comment about how meeting a new part of the farm staff (store GM) was a positive experience. Also, I briefly mentioned that some folks (not I) went to harvest potatoes, but the LLM ‘think’ once mentioning it as context to the final summary seems to have laser-focused on that topic as if it was a key theme. Anyway, take today’s AI summary with a grain of salt as usual:

AI-Generated Summary from Transcribed Field Notes

Consolidated Summary

Main Themes

1. Crop Harvesting and Preparation: Tasks like bunching chard, harvesting onions, basil, and garlic, and managing potato timing.
2. Crop Management: Monitoring crop readiness, strict timelines (e.g., potatoes), and optimizing harvest processes.
3. Collaboration and New Roles: Interaction with the new general store manager and integrating her ideas into farm operations.
4. Time Constraints: The intern’s half-day work due to a dentist appointment, highlighting operational scheduling challenges.

Key Activities

• Bunching chard, onions, basil, and garlic.
• Managing potato harvesting and immediate cooking.
• Harvesting squash.
• Engaging with the new general store manager to align operations.

New Observations/Discoveries

1. Onion Harvesting: The “green toad square” to “blue tote” conversion ratio needs optimization.
2. Basil Harvesting: Elastic-bundling technique’s effectiveness for storage or yield.
3. Garlic Volume Estimation: Challenges in estimating bunching volume during processing.
4. Potato Timing: Variable window for harvesting and cooking requires better tracking.
5. Manager’s Role: The new general store manager’s ideas and their potential impact on farm operations.

Questions for Future Research

1. How to improve the efficiency of the onion harvest-to-tote ratio?
2. Does elastic-bundling affect basil quality or yield over time?
3. What methods can standardize garlic volume estimation during bunching?
4. How to consistently track potato readiness and cooking timelines?
5. How to integrate the manager’s ideas into farm operations?

Suggested Actions

1. Refine Onion Harvesting: Track the “green toad square” to “blue tote” conversion to improve efficiency.
2. Test Basil Technique: Monitor elastic-bundling’s long-term effects on basil.
3. Standardize Garlic Volume: Use weight or visual markers for better estimation.
4. Develop Potato Schedule: Create a timeline for monitoring and immediate cooking.
5. Collaborate with Manager: Schedule regular check-ins to align store and farm goals.

This summary distills the original content into a structured, concise format, ensuring clarity and completeness.

Part 1

Main Themes

1. Crop Harvesting and Preparation: Bunching, washing, packing, and storing various crops (chard, onions, basil, garlic).
2. Crop Management: Early onion harvests, garlic bunching and drying, potato harvesting timing, and summer squash collection.
3. Collaboration and New Roles: Interaction with a new manager for the general store, highlighting her ideas and potential impact.
4. Time Constraints and Personal Notes: The intern’s half-day work due to a dentist appointment.

Activities Performed by the Intern

1. Bunching and Processing Chard:
   • Bunched 100 chard bunches, washed, packed, and fridged.
2. Onion Harvesting:
   • Gave a “lot of onions” (early onions), resulting in two blue totes after processing.
   • Bunched onions again later, with some people “munching” them.
3. Basil Harvesting:
   • Plucked basil at 2–3-inch heads, bundled with elastics, and dropped off.
   • Harvested additional basil in the dry field (though later corrected to note it was garlic).
4. Garlic Harvesting:
   • Bunched garlic, noting the volume was “a little less than yesterday.”
5. Potato and Squash Work:
   • Dug early potatoes (to be cooked immediately).
   • Harvested summer squash (limited quantity).
6. Interaction with Manager:
   • Spent time with the new general store manager, noting her ideas and potential role.

New Things Not Yet Encountered

1. Early Onion Harvesting: The process of transitioning from “green toad square” to “blue totes” after processing.
2. Basil Harvesting Technique: Using elastics to bundle basil at 2–3-inch heads.
3. Garlic Bunching and Drying: The visual and logistical challenges of estimating volume when bunching garlic.
4. Potato Timing: The strict window for harvesting and immediate cooking of early potatoes.
5. Manager’s Role: The new general store manager’s ideas and potential impact on the farm’s operations.

Questions and Future Research

1. Onion Harvesting Efficiency: How to optimize the ratio between “green toad squares” and final totes to avoid under- or over-harvesting?
2. Basil Harvesting Method: Is the elastic-bundling technique effective for long-term storage or does it affect yield?
3. Garlic Volume Estimation: How to better estimate garlic volume during bunching to avoid under- or over-processing?
4. Potato Timing: What factors influence the “variable window” for potato harvesting, and how to consistently track it?
5. Manager’s Ideas: How can the new manager’s ideas for the general store be integrated into farm operations?

Suggested Actions

1. Refine Onion Harvesting Process: Track the “green toad square” to “blue tote” conversion to improve efficiency.
2. Test Basil Harvesting Technique: Monitor if elastic-bundling affects basil quality or yield over time.
3. Improve Garlic Volume Estimation: Use a standardized method (e.g., weight or visual markers) to estimate bunching volume.
4. Develop a Potato Harvesting Schedule: Create a timeline for monitoring potato readiness and immediate cooking.
5. Collaborate with the Manager: Schedule regular check-ins to align the general store’s goals with farm operations.