Dylan Sellberg

[Dylan Sellberg @DylSell]

My favorite part about building product is that I can be told my work is "unstoppably great" and "the 667th layer of hell" in the same day

[The Neuron @theneurondaily]

AI agents are leaving the browser. 🤖🚚

@Samsara's Beyond 2026 keynote wasn't about writing emails or generating code.

It was about AI helping fleets, warehouses, job sites, and supply chains make better decisions in the physical world.

Here's what stood out. 👇

[Dylan Sellberg @DylSell]

@Mike_kim714 26.7

[Tibo @thsottiaux]

Can't wait to see what people will do with GPT-5.6 Sol Ultra. Stash your hardest prompts somewhere.

[Dylan Sellberg @DylSell]

@venturetwins @robertmclaws Same thing. I was macro tracking and told it I sometimes binge eat on weekends...suddenly it would only send an eating disorder hotline #

I bruteforced it by literally telling it exactly what I meant by binge and saying I used exaggerated language.

It's far too strict though.

[Dylan Sellberg @DylSell]

@Mike_kim714 @Titleist GTS2

[Tour Swings Tommy @TourSwingsTommy]

How I look at the flight attendant so she knows I’m ready for a treat.

[Dylan Sellberg @DylSell]

@karpathy I keep specific degradation instructions in agents.md. It is imperfect.

E.g. prioritize things from L30d ~forget things I've mentioned >6 months ago.

I think the "off the shelf" experience will get better once models remember more "trends" and fewer "facts"

[ARC Prize @arcprize]

New SOTA public submission to ARC-AGI:

- V1: 94.5%, $11.4/task
- V2: 72.9%, $38.9/task

Based on GPT 5.2, this bespoke refinement submission by @LandJohan ensembles many approaches together

[Dylan Sellberg @DylSell]

@SawyerMerritt @Tesla I don't want FSD tiers. Complicated pricing is the enemy. $99 a month flat regardless of unsupervision.

[lia @tallsnail]

love trash day. love hearing that truck. humans getting stuff done. cooperation. logistics. municipal services. civilization.

[Dylan Sellberg @DylSell]

@elonmusk When lifetime subscriptions?

[Dylan Sellberg @DylSell]

@canncon Identical experience here

[Dylan Sellberg @DylSell]

@patrickc In places with permanently posted or aerial speed traps, there's always signage about that.

And, when there are DUI checkpoints, those are also advertised.

What's the difference?

[Dylan Sellberg @DylSell]

@Codie_Sanchez Does this also = no boomerangs?

[Dylan Sellberg @DylSell]

@xleaps I've been experimenting with cursor atop obsidian on all of my docs (notes, PRDs, musings, etc.) and it looks like it just may become my default - it's amazing. Having all of that context & avoiding re-uploading each time makes a huge difference.

[Alex Veremeyenko @alex_verem]

This paper from Stanford and Harvard explains why most “agentic AI” systems feel impressive in demos and then completely fall apart in real use.

The core argument is simple and uncomfortable: agents don’t fail because they lack intelligence. They fail because they don’t adapt.

The research shows that most agents are built to execute plans, not revise them. They assume the world stays stable. Tools work as expected. Goals remain valid. Once any of that changes, the agent keeps going anyway, confidently making the wrong move over and over.

The authors draw a clear line between execution and adaptation.

Execution is following a plan.

Adaptation is noticing the plan is wrong and changing behavior mid-flight.

Most agents today only do the first.

A few key insights stood out.

Adaptation is not fine-tuning. These agents are not retrained. They adapt by monitoring outcomes, recognizing failure patterns, and updating strategies while the task is still running.

Rigid tool use is a hidden failure mode. Agents that treat tools as fixed options get stuck. Agents that can re-rank, abandon, or switch tools based on feedback perform far better.

Memory beats raw reasoning. Agents that store short, structured lessons from past successes and failures outperform agents that rely on longer chains of reasoning. Remembering what worked matters more than thinking harder.

The takeaway is blunt.

Scaling agentic AI is not about larger models or more complex prompts. It’s about systems that can detect when reality diverges from their assumptions and respond intelligently instead of pushing forward blindly.

Most “autonomous agents” today don’t adapt.
They execute.

And execution without adaptation is just automation with better marketing.

[Dylan Sellberg @DylSell]

@techNmak LLM

[Dylan Sellberg @DylSell]

@pitdesi Matic is really good. If you've got multiple floors, might be a good idea to just get 2 and get it out of the way.

[Dylan Sellberg @DylSell]

@nico_jeannen Wait 'til you hear what an engineer costs

[Jacob Posel @jacob_posel]

My favorite rebrands of 2025:

Gambling - Prediction Markets
Service Business - Full-stack startup
Sales engineers - Forward deployed engineer
PED's - Peptides/HRT
Private Equity - AI Rollup
GPU Provider - Neocloud
Docker container - RL environment
Series A - Pre-seed
Automation - AI agent
[startup] - AI [startup]