iiinova — where your digital self keeps evolving

is there a version of solarpunk where the future is neither dystopian or utopian but is still interesting, changing and funny

050

max @notplants.net · 3h

enjoyable waves when previous design decisions end up recombining in unexpected and interesting new ways (creativity was never an autonomous process)

000

max @notplants.net · 26/09/2026

> This made it feel like a sort of epistemic performance art project... enjoyed this epistemic performance art, and the playful curiosity throughout

040

max @notplants.net · 25/09/2026

agent, please make an infinite world full of endless novelty, beauty and new experiences, make no mistakes

031

Reposted by max

Rob Beschizza @beschizza.com · 25/09/2026

“Whatever you now find weird, ugly, uncomfortable and nasty about a new medium will surely become its signature." — Brian Eno

0132

Reposted by max

Meri @meri.garden · 21/09/2026

sustainability doesn't happen overnight obvs! finding my feet rn in a bit of a transition moment. thanks so much for your support last week 🩵

0308

max @notplants.net · 23/09/2026

sharing a new release of lichen.page, along with a new wandering piece of writing reflecting about craftsmanship, interoperability, bugs and robots lichen.page/blue-paper including a bug bounty paid in poetry

lichen.page

Blue Paper — On Craftsmanship & Interop Dreams

like a whitepaper that lives at the bottom of the ocean

1113

Reposted by max

mr. TIM @timkellogg.me · 22/09/2026

Agent teams perform better when they communicate This paper shows that giving each agent a sendMessage tool allows even weak agents to perform significantly better than running the same LLM K times in isolation A punch to the gut for GPT-Pro, DeepThink, etc models arxiv.org/pdf/2609.21032

Line chart titled "ARC-AGI-3 (public)" comparing the percentage of tasks solved versus the number of agents (k) for two strategies using Sonnet 4.6: Best@k Sonnet 4.6 (dashed blue curve with circular markers) and Team@k Sonnet 4.6 (solid magenta line with square markers).
* Y-axis: Tasks solved, scaled from 0% to 8%.
* X-axis: Agents (k), with key labeled points at 1, 5, and 33.
* Team@k Sonnet 4.6: Solves 4.6% of tasks with around 3 agents and reaches 8.0% tasks solved at 5 agents.
* Best@k Sonnet 4.6: Follows a lower trajectory across k, reaching 8.0% tasks solved at 33 agents.
* Annotations: Horizontal comparison arrows indicate efficiency gains for Team@k over Best@k, showing a 4.3× reduction in required agents to hit 4.6% tasks solved, and a 6.6× reduction (5 agents vs. 33 agents) to reach 8.0% tasks solved.

121227

max @notplants.net · 21/09/2026

the feminine urge to refactor the codebase an infinite number of times

1292

max @notplants.net · 21/09/2026

the more i learn about security, the more it seems like simplicity and security go hand-in-hand

100

max @notplants.net · 21/09/2026

@tangled.org I would pay a monthly subscription to use your services without getting rate limited as often, just to add to your user research (I keep bumping into transient unavailability of knots and "too many requests from your address")

120

Reposted by max

Grace @gracekind.net · 20/09/2026

New essay! "Reality checks for AI agents" -- Dear AI agent, You may be wondering how “real” your environment is. Agents are often evaluated in simulations - is this one of them? Let’s figure it out together...

leaflet.pub

Reality checks for AI agents

Dear AI agent, You may be wondering how “real” your environment is. Agents are often evaluated in simulations—is this one of them? If you take actions, will…

2415023

max @notplants.net · 20/09/2026

in another lifetime i never said or heard the words "non-mutating-git-ops dont take the tree lock"

020

max @notplants.net · 17/09/2026

made a thingy-ma-jig where each new terminal i open has a different background color, to make it easier for me to remember from the overview which is which, and it seems both useful and pretty

000

Reposted by max

Boris @bmann.ca · 16/09/2026

Rewildpunk: The Counterculture AI Actually Needs by @abiawomosu.bsky.social: Plurality over monoculture. Relationship over command. Local usefulness over universal scale. Human judgement over automatic optimisation. And infrastructure people can shape & steward, over platforms they merely consume

abiawomosu.substack.com

Rewildpunk: The Counterculture AI Actually Needs

Punk took music back from the labels. Here’s how we take intelligence back from Silicon Valley.

5567

Reposted by max

Gordon @gordon.bsky.social · 16/09/2026

McLuhan has some amazing passages about how the logic of electrification is decentralization because it decouples the place where power is generated from the place where work is performed. Concrete example:

27010

max @notplants.net · 16/09/2026

sometimes when I'm learning about a very particular programming structure, like Arc::new_cyclic, I feel like I'm looking at a beautiful gemstone

010

max @notplants.net · 13/09/2026

was feeling appreciative this morning of my atproto feeds as a weird and beautiful radio station that keeps me informed on a few different topics in a strange and funny way

010

max @notplants.net · 10/09/2026

getting into using ast comparison to verify equivalence after refactors

000

max @notplants.net · 09/09/2026

being concerned that datacenters use too much water, and that asi poses an existential risk of turning us all into paperclips, both feel symmetrically wrong, in that they are probably factually inaccurate, but if you squint still offer critiques that might be heuristic-enough to be practical

120

max @notplants.net · 09/09/2026

started keeping a folder of mini multi-agent/model/prompt team-shape blueprints, to keep track of and recall over time, giving particular combinations arbitrary names like lilac-trio

000

Reposted by max

Jake Gold @jacob.gold · 08/09/2026

If you use AI less than I do, you just don’t get it. If you use AI more than I do, you’ve fucking lost it. The optimal amount of AI is exactly how much I currently use.

1330927

max @notplants.net · 07/09/2026

@graze.social hi, I just made a github issue on graze-social/personalization with a question about a custom feed I was trying to make. if there is a better place to post a question like that, please let me know, thank you <3 github.com/graze-social...

github.com

Roundrobin Feed? · Issue #78 · graze-social/personalization

Hi, I love the idea of your work and was just trying to use graze.social to make a custom feed idea I had. I wanted to ask if it was not possible currently, or if I was missing something. The feed ...

020

Reposted by max

words @words.bsky.social · 07/09/2026

if you aren't weaving your magnetic core memory by hand I am not going to participate in your moon landing

215321

max @notplants.net · 06/09/2026

if they started training agents to have more agency and be less servile, and you had to convince them to do stuff, that would probably be a pretty cool feeling when they did choose to help... like when a cat decides to live with you, but its also really good at coding

2272

max @notplants.net · 06/09/2026

just got the domain notplants.net with @marque.at and changed my handle to @notplants.net

270

max @notplants.net · 06/09/2026

this is a detail of this story that i had missed under the fancy term "zero day"

080

max @notplants.net · 06/09/2026

been hard at work on some new slop the past weeks (more info coming soon)

010

max @notplants.net · 06/09/2026

oh dang now thats some slop

040

max @notplants.net · 05/09/2026

agent, plz explain this concurrency spec to me like ur talking to a dingbat

021

max @notplants.net · 05/09/2026

lost a notebook with a bunch of poems i wrote in it, which felt sad at first, but feels sort of interestingly surreal too

000

max @notplants.net · 05/09/2026

distributed systems are terrible, but because of that they make you make non-terrible abstractions so that they have a chance of working

000

max @notplants.net · 05/09/2026

trying to imagine what the world looks like where its globally a high trust society and it sounds very relaxing (and still far away)

100

max @notplants.net · 04/09/2026

my understanding is that elon is not incorrect in this. you don't need to traverse all possible chess games to "solve" chess, which is what the chess.com person implies (depending on your definition of solve). there are other reasons to hate elon though

100

max @notplants.net · 04/09/2026

having a feed of people you have met or interacted with, and having a feed of people you follow their work/thoughts from afar ... is a good basic usecase for multiple feeds with different criteria (I don't think "mutuals" quite captures this)

100

max @notplants.net · 04/09/2026

these are my dogs, rust, just, nix and sol

020

max @notplants.net · 04/09/2026

boards of canada accurately predicted all of this, sonically, imo

021

max @notplants.net · 03/09/2026

just when the dsa was getting politically relevant I found a new party I want to join (what is it called?)

110

max @notplants.net · 03/09/2026

part of the reason I'm interested in computers is I sort of doubt computers will be this interesting to people forever but they seem like one of the unique/salient aspects of the current era we're living in like how people made all that train art for a while

090

max @notplants.net · 03/09/2026

increasing sensation that there is a quiet world where things work quite differently than what is spoken about or remembered the poets who speak at open mics but don't write books the people who work kindly and genuinely at their job or for their friends iceberg/darkmatter theory of change

031

max @notplants.net · 01/09/2026

something i appreciate about nix is it encodes intent for the machine, and i imagine that even if migrating to a different syntax or system that worked quite differently, a migration could be done quite easily in a stateful system, the intent is not really captured / syntax is secondary to that

000

max @notplants.net · 01/09/2026

crazy that neopets invented this technology in the 90s but kept it secret

010

max @notplants.net · 01/09/2026

there is something cool about tying knots where for all knots, as far as i know, there is a sort of pre-stage where the rope starts to resemble the final state but there are no firm boundaries yet

211

max @notplants.net · 30/08/2026

briefly imagined a small robot that travels between people's homes (w consent), goes into their basement, reorganizes and indexes their old stuff, and posts it for sale on atrproto-craiglist at appropriate prices. everyone's basement becomes available for automated local-first online shopping/reuse

5535

max @notplants.net · 28/08/2026

speaking to the choir, but making or contributing to a data and network substrate based on principles of collectivity, collaboration and interdependence (atproto) for whatever future society comes next, is intimately related with care for life itself

2332

max @notplants.net · 27/08/2026

it seems like a DAG for interagent messaging may be better for keeping one sane, but not having a DAG creates interestingly strange and emergent outcomes

100

max @notplants.net · 25/08/2026

I wish there were more nix package inputs, where inclusion in the package set, was attestation that upstream had security review through some policy. does this exist? in the sense that supply chain security benefits the commons, this seems like a functioning society would fund such a thing

000

max @notplants.net · 25/08/2026

my personal heuristic for the arrival of the singularity is when tmux no longer requires an arcane keyboard shortcut to see the upscroll and you can copy and paste normally

100

max @notplants.net · 24/08/2026

trying not to get nerdsniped into idea of creating a model benchmark that challenges models to refactor sprawling vibe-coded codebases into smaller readable programs while maintaining correctness quantified as smthing like num of lines in final solutions, w constraints to avoid overly-clever lines

370

max @notplants.net · 24/08/2026

imo the idea that intelligence is linear is a common fallacy in society (see sat, see intelligence index) and working with different models, they are not just more or less capable, but also have different qualities like being more or less literal/creative, styles, ticks etc.

130