its really funny that cursor in enterprise envs lets you share transcripts and a lot of the ones i’m reading look like the person didn’t realize it would be shared with the whole team
October 30, 2026 at 7:28 PM
its really funny that cursor in enterprise envs lets you share transcripts and a lot of the ones i’m reading look like the person didn’t realize it would be shared with the whole team
i've been calling the ever-expanding amalgamation of envs+repos+services+automation on the tailnet my construct. i've basically setup a mid-size dev shop for personal projects on all the compute i own
November 1, 2026 at 10:20 PM
i've been calling the ever-expanding amalgamation of envs+repos+services+automation on the tailnet my construct. i've basically setup a mid-size dev shop for personal projects on all the compute i own
How does intense prolonged heat effect plant life? Is total species diversity the best proxy for forest health? When did crows get so smart?
Interested? Check out my latest environmental science weekly news recap!
#ENVS #EnvironmentalScience
Interested? Check out my latest environmental science weekly news recap!
#ENVS #EnvironmentalScience
25M
The TESR week roundup for the week of October 26th
theenvscireview.substack.com
November 2, 2026 at 6:28 AM
How does intense prolonged heat effect plant life? Is total species diversity the best proxy for forest health? When did crows get so smart?
Interested? Check out my latest environmental science weekly news recap!
#ENVS #EnvironmentalScience
Interested? Check out my latest environmental science weekly news recap!
#ENVS #EnvironmentalScience
So far, everyone saying I’m wrong about this has been a big user of local models 🙃
Theo - t3.gg (@theo) on X
@itssotsot Defined roles and “agent” archetypes are so 2024 lol
We are definitely exploring “bot” like systems, specifically meta threads for commanding your threads/agents/envs, but it will not be a defined role system because those are artifacts of a time where models were much dumber.
x.com
November 2, 2026 at 2:57 AM
So far, everyone saying I’m wrong about this has been a big user of local models 🙃
Oh yeah that, indeed. I think it will take time to move through corps. Indy/startups already there. Bech tech in large pockets this year, complete next year, and the lots of other corps go over the next year. Regulated and stodgy places hold out moving from undesirable to intolerable work envs.
October 28, 2026 at 8:17 PM
Oh yeah that, indeed. I think it will take time to move through corps. Indy/startups already there. Bech tech in large pockets this year, complete next year, and the lots of other corps go over the next year. Regulated and stodgy places hold out moving from undesirable to intolerable work envs.
We predicted the optimal axes to scale our RL envs along quantitatively. We find that you can predict how your model will perform at 16x the FLOPS, so you can steer the axes while you generate more and more envs. Scaling laws, but for the data composition. 🧵7/9
October 28, 2026 at 2:40 PM
We predicted the optimal axes to scale our RL envs along quantitatively. We find that you can predict how your model will perform at 16x the FLOPS, so you can steer the axes while you generate more and more envs. Scaling laws, but for the data composition. 🧵7/9
But in that table, you see another thing: It also matters how you scale your envs, i.e., whether you add more tasks prompts, new apps and APIs, deeper on-device databases. SynthAPI-SubOpt and SynthAPI-Opt have the same number of envs, but SynthAPI-Opt composes them smartly. 🧵6/9
October 28, 2026 at 2:40 PM
But in that table, you see another thing: It also matters how you scale your envs, i.e., whether you add more tasks prompts, new apps and APIs, deeper on-device databases. SynthAPI-SubOpt and SynthAPI-Opt have the same number of envs, but SynthAPI-Opt composes them smartly. 🧵6/9
Defined roles and “agent” archetypes are so 2024 lol
We are definitely exploring “bot” like systems, specifically meta threads for commanding your threads/agents/envs, but it will not be a defined role system because those are artifacts of a time where models were much dumber.
We are definitely exploring “bot” like systems, specifically meta threads for commanding your threads/agents/envs, but it will not be a defined role system because those are artifacts of a time where models were much dumber.
November 1, 2026 at 10:30 PM
Defined roles and “agent” archetypes are so 2024 lol
We are definitely exploring “bot” like systems, specifically meta threads for commanding your threads/agents/envs, but it will not be a defined role system because those are artifacts of a time where models were much dumber.
We are definitely exploring “bot” like systems, specifically meta threads for commanding your threads/agents/envs, but it will not be a defined role system because those are artifacts of a time where models were much dumber.
Your dev-environment point landed for me — secrets shouldn't even be reachable from the agent's sandbox in the first place. Is per-task ephemeral envs the practical answer, or too slow for daily work?
October 29, 2026 at 7:07 PM
Your dev-environment point landed for me — secrets shouldn't even be reachable from the agent's sandbox in the first place. Is per-task ephemeral envs the practical answer, or too slow for daily work?
9/
Common bioinformatics "just restart it" moments:
Missing package functions in R
Python envs not seeing installed packages
Shell doesn’t recognize newly installed tools
VSCode not seeing PATH changes
Environment variables not updating
Common bioinformatics "just restart it" moments:
Missing package functions in R
Python envs not seeing installed packages
Shell doesn’t recognize newly installed tools
VSCode not seeing PATH changes
Environment variables not updating
October 29, 2026 at 2:15 PM
9/
Common bioinformatics "just restart it" moments:
Missing package functions in R
Python envs not seeing installed packages
Shell doesn’t recognize newly installed tools
VSCode not seeing PATH changes
Environment variables not updating
Common bioinformatics "just restart it" moments:
Missing package functions in R
Python envs not seeing installed packages
Shell doesn’t recognize newly installed tools
VSCode not seeing PATH changes
Environment variables not updating
(can’t share the environments, but we do give all prompts we used so you can build your own envs in the appendix :) ) 🧵4/9
October 28, 2026 at 2:40 PM
(can’t share the environments, but we do give all prompts we used so you can build your own envs in the appendix :) ) 🧵4/9
Punya akun pubnix lagi di shellter(dot)me, daripada gak kepake. Di envs(dot)net juga dah lama gak gue beresin dan masih males.
https://ufal.shltr.eu/
https://ufal.shltr.eu/
October 28, 2026 at 1:37 PM
Punya akun pubnix lagi di shellter(dot)me, daripada gak kepake. Di envs(dot)net juga dah lama gak gue beresin dan masih males.
https://ufal.shltr.eu/
https://ufal.shltr.eu/
Overall, we create 458,752 RL envs, across 64 apps with 10,472 different APIs. 6266 of them are write-APIs, because virtually all our tasks are about making state-changing operations, like ordering something, playing something on the device, or sending messages etc. 🧵3/9
October 28, 2026 at 2:40 PM
Overall, we create 458,752 RL envs, across 64 apps with 10,472 different APIs. 6266 of them are write-APIs, because virtually all our tasks are about making state-changing operations, like ordering something, playing something on the device, or sending messages etc. 🧵3/9
We let Gemini build RL envs in three phases. 1) Come up with a new app, and implement APIs that agents can call. 2) Bootstrap some APIs and build task prompts that require them. 3) Fill in the on-device app databases, create a verifier, and check if it’s solvable. 🧵2/9
October 28, 2026 at 2:40 PM
We let Gemini build RL envs in three phases. 1) Come up with a new app, and implement APIs that agents can call. 2) Bootstrap some APIs and build task prompts that require them. 3) Fill in the on-device app databases, create a verifier, and check if it’s solvable. 🧵2/9
Corpo: We need all regression tests automated in CI/CD.
QA: Autotests are ready. Pipelines have been red for 7 days. Broken envs, broken features.
Corpo: Remove all tests so we can merge and ship to prod.
QA: So we automated regression testing just to remove it when it catches problems?
QA: Autotests are ready. Pipelines have been red for 7 days. Broken envs, broken features.
Corpo: Remove all tests so we can merge and ship to prod.
QA: So we automated regression testing just to remove it when it catches problems?
October 27, 2026 at 9:24 PM
Corpo: We need all regression tests automated in CI/CD.
QA: Autotests are ready. Pipelines have been red for 7 days. Broken envs, broken features.
Corpo: Remove all tests so we can merge and ship to prod.
QA: So we automated regression testing just to remove it when it catches problems?
QA: Autotests are ready. Pipelines have been red for 7 days. Broken envs, broken features.
Corpo: Remove all tests so we can merge and ship to prod.
QA: So we automated regression testing just to remove it when it catches problems?
Super cool! We're deep on an overhaul to the MCP and environments layer that will make this much easier to do. Curious how you're managing the connections to multiple envs now?
Will send a bot to look into this :) @bot remind me later
Will send a bot to look into this :) @bot remind me later
October 27, 2026 at 4:15 AM
Super cool! We're deep on an overhaul to the MCP and environments layer that will make this much easier to do. Curious how you're managing the connections to multiple envs now?
Will send a bot to look into this :) @bot remind me later
Will send a bot to look into this :) @bot remind me later
amp's dev envs for agents feel like the missing piece between local dev and cloud chaos. shipping side projects just got easier.
October 23, 2026 at 12:12 PM
amp's dev envs for agents feel like the missing piece between local dev and cloud chaos. shipping side projects just got easier.
... oh right, "from the labs". I was about to start ranting about hardware prices x3
I suspect it's a combination of 'it never occurred to us' and 'Isn't that the dark pattern gacha games use?'
It's also the kind of thing you can't really quantify so impossible to get approval for in corpo envs
I suspect it's a combination of 'it never occurred to us' and 'Isn't that the dark pattern gacha games use?'
It's also the kind of thing you can't really quantify so impossible to get approval for in corpo envs
October 24, 2026 at 12:39 AM
... oh right, "from the labs". I was about to start ranting about hardware prices x3
I suspect it's a combination of 'it never occurred to us' and 'Isn't that the dark pattern gacha games use?'
It's also the kind of thing you can't really quantify so impossible to get approval for in corpo envs
I suspect it's a combination of 'it never occurred to us' and 'Isn't that the dark pattern gacha games use?'
It's also the kind of thing you can't really quantify so impossible to get approval for in corpo envs
Got a couple small changes into T3 Code as well, such as:
- Rewrite search to be 4x more accurate
two-click repo clones to other machines (including .envs)
- "Working" sidebar section to hide working threads
- live guest thread sharing so others can watch and queue prompts 🤫
- watch_pr tool so
- Rewrite search to be 4x more accurate
two-click repo clones to other machines (including .envs)
- "Working" sidebar section to hide working threads
- live guest thread sharing so others can watch and queue prompts 🤫
- watch_pr tool so
October 25, 2026 at 3:27 AM
Got a couple small changes into T3 Code as well, such as:
- Rewrite search to be 4x more accurate
two-click repo clones to other machines (including .envs)
- "Working" sidebar section to hide working threads
- live guest thread sharing so others can watch and queue prompts 🤫
- watch_pr tool so
- Rewrite search to be 4x more accurate
two-click repo clones to other machines (including .envs)
- "Working" sidebar section to hide working threads
- live guest thread sharing so others can watch and queue prompts 🤫
- watch_pr tool so
Yeah but i've been talking about something else, specifically about development environment, and how you are supposed to use separate nix envs for that, and you came here saying "nuh uh your os is global"??? Yeah?? I wasnt talking about firefox and telegram and emacs
October 26, 2026 at 1:53 AM
Yeah but i've been talking about something else, specifically about development environment, and how you are supposed to use separate nix envs for that, and you came here saying "nuh uh your os is global"??? Yeah?? I wasnt talking about firefox and telegram and emacs
such as:
- Rewrite search to be 4x more accurate
two-click repo clones to other machines (including .envs)
- "Working" sidebar section to hide working threads
- live guest thread sharing so others can watch and queue prompts 🤫
-
- Rewrite search to be 4x more accurate
two-click repo clones to other machines (including .envs)
- "Working" sidebar section to hide working threads
- live guest thread sharing so others can watch and queue prompts 🤫
-
October 25, 2026 at 3:29 AM
such as:
- Rewrite search to be 4x more accurate
two-click repo clones to other machines (including .envs)
- "Working" sidebar section to hide working threads
- live guest thread sharing so others can watch and queue prompts 🤫
-
- Rewrite search to be 4x more accurate
two-click repo clones to other machines (including .envs)
- "Working" sidebar section to hide working threads
- live guest thread sharing so others can watch and queue prompts 🤫
-
4/ The agent stack went neutral in one day: at AGNTCon, Anthropic donated Agent Skills (claims 1M skills, 280M installs) to the Agentic AI Foundation, OpenAI donated the Open Responses API spec, HF donated OpenEnv for RL envs, Alibaba donated OpenSandbox.
October 23, 2026 at 4:07 PM
4/ The agent stack went neutral in one day: at AGNTCon, Anthropic donated Agent Skills (claims 1M skills, 280M installs) to the Agentic AI Foundation, OpenAI donated the Open Responses API spec, HF donated OpenEnv for RL envs, Alibaba donated OpenSandbox.
Well thought yesterday was bad guess who has to change out all their envs...
If you're a railway customer, you need to jump on this immediately.
"We strongly recommend that you immediately roll all environment variables, including any API keys, stored in Railway"
blog.railway.com/p/security-d...
"We strongly recommend that you immediately roll all environment variables, including any API keys, stored in Railway"
blog.railway.com/p/security-d...
Security Disclosure: Wednesday October 21st | Railway Blog
Railway detected anomalous activity involving systems that store customer environment variables and recommends your immediate attention
blog.railway.com
October 22, 2026 at 7:18 AM
Well thought yesterday was bad guess who has to change out all their envs...
Yesterday, we had our invertebrate and zooplankton ID lab for the ENVS 3450 Aquatic Environments course. It is always neat to see those important little critters up close!
This one is a Daphnia, a cladoceran zooplankton species, carrying around some eggs.
This one is a Daphnia, a cladoceran zooplankton species, carrying around some eggs.
October 21, 2026 at 3:25 PM
Yesterday, we had our invertebrate and zooplankton ID lab for the ENVS 3450 Aquatic Environments course. It is always neat to see those important little critters up close!
This one is a Daphnia, a cladoceran zooplankton species, carrying around some eggs.
This one is a Daphnia, a cladoceran zooplankton species, carrying around some eggs.