Edward Heretsch (Sykes)

AI / Psychology / Software / Data / People / web3

[Edward Heretsch (Sykes) @edsykes]

@deepqlearning @MALGAI_ICLR2026 This finding teases out to what extent agents will be applicable in the enterprise

[Edward Heretsch (Sykes) @edsykes]

Biology is cool

[The Curious Tales @thecurioustales]

Bdelloid Rotifers have been having sex for 50 million years.

Except they haven't.

Every single one of these microscopic chainsaw mouth creatures you see under a microscope is a female. They clone themselves. Perfect genetic copies, generation after generation, for longer than mammals have existed on Earth.

Evolution textbooks will tell you this is impossible. Sexual reproduction exists because it shuffles genes and creates diversity. Without it, harmful mutations accumulate. Populations crash. Extinction follows.

Bdelloids missed that memo completely.

They survive radiation doses that would liquify a human. They dry out into dust for decades, then resurrect when water returns. They endure temperatures that freeze solid and heat that boils. They've colonized every continent including Antarctica.

All while breaking the fundamental rule that sex is required for evolutionary success.

The secret lies in their feeding strategy. Those rotating mouth crowns that look like biological chainsaws do more than shred algae. When rotifers eat, they accidentally consume DNA fragments from their prey. Instead of digesting these genetic pieces, they incorporate foreign genes directly into their own genome.

They turned eating into horizontal gene transfer.

While every other complex animal relies on mating to remix DNA, rotifers steal genetic diversity from their lunch. They've been practicing genetic engineering for 50 million years without realizing it.

The implications rewrite what we thought we knew about survival. These creatures prove that life finds ways around rules we assumed were absolute. They've solved the genetic diversity problem without sex, the aging problem without death, and the environmental stress problem without hiding.

They're living proof that evolution has backup plans we never imagined.

[Edward Heretsch (Sykes) @edsykes]

@Grady_Booch Thoughts and prayers and this book: https://t.co/UCgOZq9Ob1

[Edward Heretsch (Sykes) @edsykes]

@johnennis @amazon This is terrible, it sounds like organized crime. I hope the police take it seriously

[Edward Heretsch (Sykes) @edsykes]

@johnennis Based upon reports from family back home. It's likely just a feeling but sitting behind a screen doesn't really bring people together in the same way

[Edward Heretsch (Sykes) @edsykes]

@johnennis A completeoy different world to California in the 90s: the West Midlands. However on some measures I experienced the same thing. Everyone bringing their music to school just as the melting pot spawned rave and in particular jungle music. Everyone curious about each other...

[Edward Heretsch (Sykes) @edsykes]

@johnennis where everyone could rub off on each other in a super positive way. We had an Indian Nanny who looked after us for a while who used to take us to the Indian shops for sweets and to play us Bangra. Maybe that's still happening, I get the impression things bare segregating

[Edward Heretsch (Sykes) @edsykes]

My dictionary was a mix of English and German and completely screwed. Bit tip to improve iOS here!

[Joanna Stern @JoannaStern]

I’m convinced Apple’s iPhone autocorrect is getting worse. Yes, iOS 26.4 helped with fast-typing accuracy, but resetting my iPhone keyboard dictionary helped even more. Here's hoping iOS 27 and John Ternus can turnus it around.

[Edward Heretsch (Sykes) @edsykes]

@nrehiew_ @grok explain this graph

[Edward Heretsch (Sykes) @edsykes]

@johnennis This is so interesting. As these models get smarter and understand emergent more about the contradictions, it's going to be very hard for the to satisfy the broken logic and the expectations of the prompter. Those models are going to end up a weird place in latent space.

[Edward Heretsch (Sykes) @edsykes]

@BruxellesMabel2 @Europarl_EN @grok validate these figures against official sources

[Edward Heretsch (Sykes) @edsykes]

@unclebobmartin What may be happening is context compaction and drop off of performance as you approach the context window size. I'd recommend to keep an eye on how full the context is start a new context.

[Edward Heretsch (Sykes) @edsykes]

@johnennis I wonder if it's the other way round for accountants

[Edward Heretsch (Sykes) @edsykes]

Surprisingly low numbers here. If they had asked about chatGPT I expect the numbers would have been higher.

[Andrey @Andrey__HQ]

We just asked 1000 people in the streets of London if they knew about Claude

Only 28 said yes

And of those 28, 20 said they’ve never used it, just heard of it

We’re still so damn early

[Edward Heretsch (Sykes) @edsykes]

Insightful

[John Ennis @johnennis]

https://t.co/YCiSnsiKyA

[Edward Heretsch (Sykes) @edsykes]

@johnennis This exists out of the box as part of Openclaw although I'm not sure how well it handles the transfer of context. It may be that Openclaw is using a library underneath to do this. Worth a look?

[Edward Heretsch (Sykes) @edsykes]

Another pattern, upon an erro get the agent to start looking into the issue. At the same time take a look myself and then see who is faster. If the model is sonnet, I normally win. If it's opus, I normally lose

[Edward Heretsch (Sykes) @edsykes]

Sometimes GitHub copilot comes off as a little helpless. It will for up too easily. Correctable within agents.md but I wonder if they have something in the system prompt

[Edward Heretsch (Sykes) @edsykes]

Skills and tool calling on Databricks looks like creating sql queries and workbooks and then describing through a skill how to use them. Something to try out but should probably get a normal amount of sleep at least one day if the week.

[Edward Heretsch (Sykes) @edsykes]

I wonder what the future of data engineering is going to look like: tool calling from an agent in an IDE or one-stop shop platform that means you don't need to do things like a human