Even before the agentic revolution, prompting tricks stopped being very valuable, as our research has shown. The format isn't key, the best approach to AI right now is to clearly specify your goals, your output, what "good" & bad look like, how to test the results... (yes, this is just management)

Cloud computing gives us more ways than ever to work with genomic data, but each has its own intricacies. Maintaining that storage can dwarf the cost of using or renting it, yet optimizing it slips down the backlog in favor of quick fixes. hubs.la/Q04nzSl40

Found a company selling "whole genome sequencing - 30X" for any species, human or pet (www.the-odin.com/whole-genome...). Also, I am now very tempted to adopt an Australian lungfish as a pet.

on this fourth of july i want to reinforce my belief that i think it's nuts that entrepreneur culture ever forced us to feel like a bunch of dullard MBA guys in the right place at the right time were ever worthy of mass praise and adulation.

Also interesting to see moves in the opposite direction where a cheaper model can tag in a smarter model: platform.claude.com/docs/en/agen...

What if the model is the router? I think people underestimate the ability of frontier models now, but especially in the near future, to delegate work on their own as needed to dumber, cheaper models. The right architecture may be “start with a smart AI planner, let it delegate on its own”

You know what just occurred to me? LLMs write interview code. Think about every coding interview you’ve done or conducted: it has this “show your work” look with very little design (no time) very little use of libraries and so on. Now go look at your baseline LLM code. Yup. Same thing.

Great Success! Finally submitted something to fable and didn't have it immediately tell me to GTFO. For all I know it's silently downgraded me to Opus, but it's progress! 😅

Laugh all you want, but what is your explanation for the Paul George for Jaylen Brown trade?

I realized today that while I'd been using "robot overlord" as my tongue in cheek metaphor for AI tools, that a better one would the the Brains from Futurama

I only quickly skimmed the other archetypes at bambamramfan.github.io/ai-compass/ but I find it hard to believe another one would capture my persona better than what it assigned me.

Really jinxed that one

Putting AI to good use!

I’m often even more general and just say there’s only one job: solving problems and getting shit done. Everything else is tooling.

Two ICE agents tracked down a woman *inside a polling place* yesterday. Why? They demanded that she take down an Instagram post in which she named Jonathan Ross as the man who shot & killed Renee Good (after the Star Tribune identified Ross). www.syracuse.com/news/2026/06...

I'm collecting stories of bioinformatics where LLMs failed. If you have examples of attempts to use LLMs for analyses or pipelines that failed, please share. I want successes too; if I get an anonymized collection I'll publish it.

What are the chances paying this ambulance bill w/ my credit card gets coded as "travel" so I get the bonus points? If I were in charge I'd set it up even if just for the lulz

@unlikelywords.bsky.social got a double compliment on the gracies op ivy shirt today in inman. They said they loved the shirt, and also that y'all make the best shirts.

So barrel aged malort is a wee bit smoother. But to what end

can't argue with this.

I have a skill that checks my prose for AI tells. It mostly just flags vocal tics that I've had my entire life. I guess I've been a bot all along?

I took one look at that man’s hair and decided to trust everything he had to say about pool and pool related products

love how every app has to periodically inform me that push notifications are disabled. damn bro who do you think did that!!! history's mysteries

@willgordonagain.bsky.social I assume you've gone to at least one of these cambridge watch parties? If so - is the setup worth it compared to going to a local watering hole?

ngl it legit cracks me up when Claude gets on that thing that's basically it saying "Hey, I would like to stop talking about this now. Please move on"

Minibwa is a hybrid of bwa-mem and minimap2 and the successor of bwa-mem for short-read mapping. ~4X/2.5X as fast as bwa-mem/bwa-mem2 for WGS reads at comparable accuracy. Native support of directional bisulfite-seq. Applicable to long reads. Preprint at arxiv.org/abs/2606.15357

Oh my god I hope that the @redsox.com know that NASA just announced the selection of a new mission called GOSoX out of MIT!!!

Every time this telemundo announcer says something about the guy Summerville on the NL team it really throws me for a loop.

“Scots discovering tailgating” could be a real Prometheus stealing fire moment

I'm pretty happy that I type "hambu" into Youtube and the 2nd hit was "Hamburger Lady"

Quote from a Claude session ❤️ "and it's the moment you swore at me" Claude, this just means you're part of my inner circle now. You'll get used to it.

The Economist out here dropping hard truths on me before my second coffee this morning.

you can just say things

OCEANIA HAS ALWAYS BEEN VERY CLOSE TO A DEAL WITH EURASIA

no one has ever earned gold or experience points from a status update meeting

Every damn morning i freak out as all of a sudden my slack notification count shoots up. "Uh oh, did something break?" "what's wrong?" etc. No, it's 9am and all of my "save until tomorrow" clicks just fired. 😅 One of these days that'll stop triggering my fight or flight, but not today.

Fed claude somef things I've written & asked it to analyze my voice. It told me I think in analogies & abstract systems, and am comfortable with ambiguity. And that I overuse self-effacing language & parentheticals. (have been working on the last one recently!) Claude gets me! It really gets me!

We’re doing a user study to find out how adding always-on live values alongside code effects writing and debugging experiences for functional programming. See reply for sign-up details

I'm alive *AND* I wrote a blog post?! Unheard of! Much 💖 to @abbycabs.dev, and @grafana.bsky.social's David Allen and Amanda Wagner whose FOSDEM talk's were a huge source of inspo!

Saw a Raspberry Pi project called "Doodle Detector" and it took me a moment to realize it wasn't the equivalent of Silicon Valley's "Not Hotdog"

My least favorite vibe coding pattern is people raw dogging the database.

Earlier today I noticed Claude going off the rails. I hit escape to stop it so I could course correct. It immediately started back up again. I saw it had left a message "I'll treat the skip as 'go ahead'". Not what I meant there, buddy

No, Papa John's app. I, and I cannot express this strongly enough, do not want your AI assistant to decide what to put on my pizza. What are we even doing here?

Am I misremembering or does Murderbot use more parentheticals as the series progresses? (I’m probably misremembering)

Say what you will about Caligula, but he appointed an *entire* horse to a government position, not just the rear end.

When people say 'AI isn't a bubble because the technology isn't going away', it just shows they don't know what a bubble is. The internet didn't go away. Railways didn't go away. Tulips didn't go away. An asset can be overvalued, even if its useful.

Benedict Evans was on Lenny’s podcast and reminded me of a very useful dichotomy: “task vs. job.” I think folks who define their career in terms of tasks have a very different mindset, especially about AI, than do folks who define their jobs as something high level (not specific tasks).

lol, went from thinking it was a local thing to seeing people in upstate NY reporting seeing a fireball in the sky. Yeesh.

FYI