I watched this while feeding my infant daughter and reflected on how many lies I’ve already told her.

I ran this one while listening to music but still trying to keep the 160spm cadence, based on the watch metronome vibrations. Went just ok.

Unknown location (probably inside)

Maybe this movie is intentionally unintelligible. Maybe the cliches are perfectly chosen and not ineptly plopped on a page. Maybe the bad guy’s lack of motive was the really fearful part. Maybe the girls’ side plot was really relevant as contrast to no plot in the main plot. Maybe the disorientation of sloppily shot sequences and non-sequitur statements were meant to convey the doubt and feeling of Dakota’s character. 

No, she went to the eye doctor because she could see the future. This was stupid all the way through.

The software engineering industry has had a standard of adding a README.md file to the root of projects and also the root of any subdirectory of those projects where deemed necessary. These files are easily discoverable by operators of the codebase, and their contents - while not standard - have been embraced as a place to put instructions for patterns and how to operate in the codebase.

with the rise of coding agents over the last couple years, people found that they needed to give them special instructions because they were error prone in certain ways, and we hadn’t built out the harnesses and capabilities for them to replicate how human operators work in a code base. That is no longer true today.

We We have incredibly capable models and the harnesses and tooling that we give them (like shell access and MCPs for browser control, among hundreds of other tools), being that they can do everything that a human operator can do to operate on the code. Everything I would say to a human operator in the codebase, I would say to an LLM agent working in that same codebase.

So, I just revamped my project’s READMEs and symlinked them to the AGENTS.md location. It’s not useful to separate the instructions for humans from LLM agents. In fact, when we separate them, we increase the likelihood that they will operate in different ways and do things that the other does not expect or intend. This actively slows down development on both sides.

Solidifying a single place - the old README.md standard, that is present in all modern software - is the path forward. I’m symlinking my READMEs to conform to this standard for now, because it’s free and doesn’t clutter anything for me, but I hope it eventually falls away.

Most times I get on this treadmill I’m shown the path by our rental apartment in Auckland, New Zealand and I miss being in that sunshine.

Unknown location (probably inside)

Taken from Tom Sachs.

If the patient asks, “Will this hurt?” [music] Always reply with, “Not as much as the thing you did to get yourself in this situation.

FROM:

Tom Sachs

Knot Tying Tutorial

Source

But more seriously, this is also being converged on with things like Shopify’s roast framework or Karpathy’s autoresearch framework. The important part here is using a meta-harness to instruct the agents on constraints, stitch them up with deterministic programs, etc.

I’m continually impressed with Claude’s ability to create its own benchmarking systems and rip through a series of hypotheses about how to achieve some hard-to-define optimal outcome.

I recommend trying this when you’re deep into a problem looking for an optimal path: give the agent a laboratory.

FROM:

Writing

Give Your Agent a Laboratory, Pt. II

Source

This is another person discovering Karpathy’s autoresearch: deceptively simple, but give an agent a goal and space to pursue it and you will get improvements.

This was easier or harder than what my daughter made it through in tummy time?

Unknown location (probably inside)

I have watched death as it happens, and I have watched birth. What I learned is that neither are discrete events. They are both processes, things that unfold. Birth is a series of awakenings, and death is a series of sleepenings. My son will take years to be born, and my father took six months to die. Some people spend decades dying.

FROM:

Dean W. Ball

Clawed

Source

Even if Secretary Hegseth backs down and narrows his extremely broad threat against Anthropic, great damage has been done. Even in the narrowest supply-chain risk designation, the government has still said that they will treat you like a foreign adversary—indeed, they will treat you in some ways worse than a foreign adversary—simply for refusing to capitulate to their terms of business. Simply for having different ideas, expressing those ideas in speech, and actualizing that speech in decisions about how to deploy and not deploy one’s property. Each of these things is fundamental to our republic, and each was assaulted—not anything like for the first time but nonetheless in novel ways—by the Department of War last week. Most corporations, political actors, and others will have to operate under the assumption that the logic of the tribe will now reign.

There is something deeper about the damage done by the government, too. The Anthropic-DoW skirmish is the first major public debate that is truly about where the proper locus of control over frontier AI should be. Our public institutions behaved erratically, maliciously, and without strategic clarity. Our political leaders conveyed little understanding of their own actions, to say nothing of the technology and its stakes. They got off on an extraordinarily bad footing, and it is hard to imagine them ever recovering, because they do not seem to care about improvement. They are a cartoonish depiction of the American political elite, but sadly their failings have been the prototype of American political elites from both parties for much of my life now. “The same as before, but now noticeably worse” has been the theme of American politics for 20 years.

FROM:

Dean W. Ball

Clawed

Source

This is a great and measured piece about how terribly stupidly bad it was to do what they did last week.

Ran to the ballot box to drop off our state primary mail-in ballots at the library and all the polling place old ladies were happy to help me.

On the first nice spring day, everyone is happy.

Oh! And I successfully ran a 160spm cadence just with the pulse from my watch. No music required.

Chicago, Illinois, United States (Weed 2 Halsted)

I think Dance Yrself Clean might be the best drop in all of concert music. The bass, the falsetto synth, the minimal screams. Every time it hits the whole crowd goes crazy.

Automatic was their opener and I went halfway through their set before realizing that I had listened to their album Is It Now? a lot last year. What a lovely surprise.

Leg day was scheduled and I almost changed it for something more fun but then I reminded myself that it all can be fun and so did it anyway.

Unknown location (probably inside)

This one popped into my head after seeing some coworkers blindly trusting an LLM/AI agent’s output, which was entirely incorrect.

This is my little reminder to always check the output of today’s AI agents. Know the checks applied to their work before it reaches us. Understand the provenance of their output. Their incentives are not your own.

It was a long day in the software mines so I enjoyed listening to Zvi’s podcast recount the industry back to me with snark while I lifted weights.

Unknown location (probably inside)

But that was very hard to follow, surprisingly! I had to stop the podcast and switch back to the music to keep my cadence. I’m still learning.

Unknown location (probably inside)

I brought them a jade plant as thanks for wonderful treatment.

Unknown location (probably inside)

The sole objective today was to increase my cadence by 5% and run at 160spm (reasoning here). I found a playlist of popular songs mixed to 160bpm and listened to the inanity throughout. I was prepared to see a slower time, expecting my stride to compress, but apparently my body has a stride length constant and the faster steps just increased my speed overall by like 5%. What a trick! I don’t even feel more exhausted.

I think the ideal 160spm song might be Shake It Off?

Chicago, Illinois, United States (Halsted to Larabee)

Same impulse that led me to build iCal feeds for my entire blog history. A calendar is the tool I already use for planning around time. Price data is time data — it just happens to come from a utility instead of a CMS. If I can see that electricity drops to near-zero at 2am and spikes at 6pm, I can plan around it the same way I plan around meetings.

So I built a small Val.town server that generates an iCal feed of price changes. It pulls the last 24 hours of 5-minute prices from ComEd’s public API, averages them into hourly buckets, and grabs the next day’s prices from their (undocumented) day-ahead endpoint. Then it compares consecutive hours and emits a calendar event whenever the price shifts by more than a threshold:

↑ 3.7c/kWh (+0.9c)
↓ 2.1c/kWh (-0.3c)

Stable hours produce no event — gaps in the calendar mean the price isn’t moving. The sensitivity, lookback window, and lookahead are all configurable via query parameters, so I (or you!) can tune it to only surface the swings I care about.

Note

Day-ahead prices aren’t always available — ComEd publishes them on their own schedule, typically in the evening for the following day — so the feed only includes forward-looking events when the data is there.

The next step is pairing this with batteries to buffer my high-draw appliances — grow lights, the computer desk — into cheap hours automatically. For now, just seeing the price rhythm on my calendar alongside everything else is enough to shift my habits. Everywhere a calendar.

Note

They expose this data over a decent API. I should use this along with batteries to make cheap-electricity battery-buffers for my high-usage appliances (e.g. LED grow lights, computer desk, etc.).

If whatever I was doing on the kitchen counter is now called “software engineering,” then ordering food at a restaurant should be called “cooking.” As much as I marvel in this new and (dare I say) magical way of manifesting products and services from thin air, I question whether it is truly a creative process anymore. Inasmuch as we pursue craftsmanship as a goal unto itself, what’s the point for us humans when the machines are going to be better, faster, and cheaper than all of us?

FROM:

Ben Sigelman

The Giddy Nothingness of Automatic Creation

Source

If you didn’t make decisions, you weren’t engineering. The engineering - the value you bring - and work is in judgement. You need to decide the trade-off to take and the checks to put in place to let the system operate. You are now building and operating a system that builds the actual code. Your value is in knowing how that system will fail, or diagnosing and discovering failure, and preventing that (while ensuring the actual goal is achieved). We always had to do that in the before times, but now the code writing is not required of you.

I’m in the gym feeling good and then in the TV I see that Trump illegally fired missiles (with Israel) into Iran to kill people, without consent of Congress, for no stated reason or imminent danger. This, while I’m listening to a podcast with Attorneys General talking about the illegal use of federal power to compromise states’ rights inside the US.

Unknown location (probably inside)