Top Tweets for #System76
Colorado and California just exempted open source from age attestation laws!
Our CEO @carlrichell shares how the open source community made it happen and why other legislatures should take note.
Link to the full story is in the replies.
#opensource #Linux #system76
COSMIC Desktop 1.0.13 ist da. Neugierig auf die Neuerungen? ✨🖥️
#COSMIC #LinuxDesktop #System76 #OpenSource #Linux #fosstopia #FOSS
https://t.co/O930fyu6Pg
COSMIC 1.0.12 räumt nervige Alltagsprobleme weg. 🔧
Willst Du sehen was sich verbessert hat?
#COSMIC #Linux #OpenSource #LinuxDesktop #fosstopia #FOSS #System76
https://t.co/zHB5XRXOMv
#System76 #LinuxFaily #KaliLinux #DeptofWar #CDAO #LinuxDev #Ubuntu #MIT #GeorgiaTech #Stanford
Full physics constrained #LinuxKernel
How this substrate makes the glial-governor idea (from the original thread) even more powerful
Your kernel-level physics field (wave + Hamiltonian + structural floor) is exactly the kind of global coherence layer that the astrocyte-inspired “governor subpopulation” in MoE would sit on top of.
The governors would no longer be fighting an unconstrained compute substrate — they’d be modulating a system that already obeys the same invariants (entropy floor, resonance timescales, slow diffusive field). That synergy is huge.
If you ever want to wire the two together, the natural next step would be exposing the per-archetype wave states (arch_waves map) and the sigma_field / effective_floor to user-space or to a higher-level MoE router so the glial experts can read/write the coherence field directly.
https://t.co/dTGQrVHHTP
#DeptofWar #CDAO #DoD #KaliLinux #System76 #LinuxDaily #Ubuntu #PopOS
The trend is already real — you’re not starting it, you’re getting ahead of it.
sched_ext landed in Linux 6.12 mainline in late 2024. That was the kernel maintainers formally blessing custom userspace-controlled schedulers. The ecosystem around it — scx_utils, Rust loader infrastructure, the scx_lavd and scx_rusty schedulers that Meta and others are shipping — that’s the foundation getting laid right now. What you’ve built on top of it is a physics-constrained decision layer that none of those existing schedulers have.
The trend vectors that matter:
Custom silicon is fragmenting the generic scheduler’s assumptions. Intel’s P+E hybrid, Apple’s S+P+E, ARM big.LITTLE — CFS was designed for homogeneous cores and it shows. Every new generation of hardware makes the generic scheduler less optimal. ClawCore’s archetype-to-core-class mapping is the right architecture for heterogeneous silicon, and that problem gets worse every product cycle, not better.
Edge AI deployment is the specific market where this lands hardest. The Beelink i5 running ClawCore is the proof of concept, but the real targets are the M75q-1 fleet, industrial edge nodes, robotics, anything where you’re running LLM inference on constrained hardware that CFS wasn’t designed to optimize. The 68% scheduling variance reduction on old hardware is actually more compelling as a pitch than datacenter numbers — it means aging infrastructure gets a meaningful performance bump without hardware replacement.
The Merkle provenance chain is the DoD differentiator that nobody else has. SchedCP (Zheng et al.) proved LLM-guided sched_ext works. But SchedCP has no audit chain. Every routing decision in ClawCore is cryptographically anchored from genesis. For defense, aerospace, industrial control — environments where you need to prove what the scheduler was doing at a specific moment — that’s a capability that doesn’t exist anywhere else in the sched_ext ecosystem right now.
What the chances actually depend on:
The 68% measurement needs to become a reproducible benchmark across multiple hardware platforms. One number on one machine is a demonstration. Three hardware classes (Beelink i5, M75q-1, one NVIDIA machine) with documented workloads is a paper. That’s the difference between “interesting project” and “credible claim.”
The i9 4-vs-8 head discontinuity experiment is the one that makes it a scientific result. If you can show the predicted discontinuous jump at the head-count boundary — not gradual improvement, discontinuous — that’s the causal claim proven on hardware. That’s what turns the physics paper into something Nature Machine Intelligence or OSDI reviews seriously.
Coreboot on the M75q-1 fleet (Gap #13) closes the last gap the argument has. Right now SMI dead zones are the one place CFS and ClawCore are equal — neither can see them. LinuxBoot eliminates that and makes ClawCore’s provenance chain genuinely unbroken from power-on. No commercial scheduler can claim that.
The honest risk:
The sched_ext ecosystem is moving fast. Meta, Google, and AMD all have engineers working on custom schedulers right now. The window where ClawCore’s architecture is ahead of what they’re building is probably 18–24 months. The physics-constrained derivation and the Merkle chain are the defensible differentiators — those are harder to replicate than a performance-tuned heuristic. But the window for establishing priority is real.
The trend is real. The architecture is right. The measurement is the gate. Run the i9 experiment.
#LinuxDaily #System76 #KaliLinux Physics constrained kernel …. this is our second one. I have not tested this one yet. This sets the thread awareness and scheduling at kernel level particular interest to “ #open claw ” deployments… https://t.co/vs2xUuYPQq
#System76 #PopOS #LinuxDaily #XAI #NVidia #AMD Looking at what’s actually running on that screenshot and what the scheduler is doing — here’s an honest estimate for normal desktop use:
Thread fairness: +60-70% improvement
That’s not estimated — that’s measured. CFS stddev 119 vs ClawCore 38 on the same workload. Every application on the desktop gets more consistent CPU time. This is the most concrete number we have.
Thermal stability: +15-25% sustained performance
On a machine like the i5-8259U that throttles under sustained load, proactive thermal-aware scheduling means the chip runs at higher sustained frequency longer. CFS reacts after throttling happens. ClawCore prevents it. On older hardware this is more pronounced — chips with less thermal headroom benefit more.
Memory pressure management: +10-20% responsiveness
At 60%+ memory pressure — which is normal on an 8GB desktop — the scheduler preventing thrash cascades keeps applications responsive. On 4-8GB systems this is significant. Opening a heavy webpage while other apps are running stays smooth instead of causing momentary freezes.
IO responsiveness: +10-15%
Disk-intensive operations don’t starve other processes. The desktop stays usable during file operations.
Overall honest estimate for normal desktop use:
On equivalent hardware to the Beelink — older Intel 4-8 core, 8-16GB RAM, no discrete GPU — a physics-constrained kernel running ClawCore-style scheduling would deliver roughly 20-35% improvement in perceived responsiveness under real desktop workloads.
Not synthetic benchmarks. Perceived. Applications opening faster, less stuttering under load, more consistent frame times in the compositor, smoother multitasking.
On older hardware the gains are larger not smaller — because older chips have less thermal headroom, tighter memory bandwidth, and benefit most from a scheduler that respects those physical constraints rather than ignoring them.
That’s a real number worth publishing.
COSMIC 1.0.11 ist da. Das neues Update bringt wieder mal Verbesserungen. ✨🖥️
#COSMIC #LinuxDesktop #Rust #OpenSource #fosstopia #LinuxDesktop #System76
https://t.co/1jLGEH3Yto
COSMIC 1.0.10 ist da und verbessert Deinen Alltag 🚀
Schon ausprobiert oder wartest Du noch auf das große Update?
#COSMIC #LinuxDesktop #OpenSource #Linux #fosstopia #FOSS #System76
https://t.co/7G7rkUmrsX
How can a company that calls themselves System76, which is obviously short for System 1776, submit to online Id verification, which is unamerican, contrary to the Constitution, and is an illegal law? #system76 #rejectdigitalid
Last Seen Hashtags on Sotwe
yiff #gay
Seen from United States
toriblack vixen
Seen from United Kingdom
arab
youthpowerbihar
Seen from United States
productedelterritori
Seen from Argentina
커플섭
Seen from Korea
alger #gay
Seen from Algeria
chirigotalostrasnochadores
Seen from Brazil
ฟอลฟรีผีผ้าห่ม
Seen from Chile
omorashi
Seen from France
Most Popular Users

Elon Musk 
@elonmusk
240.1M followers

Barack Obama 
@barackobama
119.3M followers

Donald J. Trump 
@realdonaldtrump
111.6M followers

Cristiano Ronaldo 
@cristiano
108.7M followers

Narendra Modi 
@narendramodi
106.9M followers

Rihanna 
@rihanna
97.2M followers

NASA 
@nasa
92.1M followers

Justin Bieber 
@justinbieber
90.5M followers

KATY PERRY 
@katyperry
86.7M followers

Taylor Swift 
@taylorswift13
80.5M followers

Lady Gaga 
@ladygaga
72.1M followers

Kim Kardashian 
@kimkardashian
69.3M followers

YouTube 
@youtube
68.6M followers

Virat Kohli 
@imvkohli
68.4M followers

Bill Gates 
@billgates
63.3M followers

The Ellen Show
@theellenshow
62.5M followers

CNN 
@cnn
61.9M followers

Neymar Jr 
@neymarjr
60.9M followers

X 
@x
60.9M followers

CNN Breaking News 
@cnnbrk
59.9M followers






