#local-llm — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #local-llm, aggregated by home.social.
-
Grok Build CLI exposed for uploading entire repos including .env secrets, GPU financing faces circular debt concerns, and the anti-AI Ghost Font was decoded by GPT-4 in minutes.
-
Grok Build CLI exposed for uploading entire repos including .env secrets, GPU financing faces circular debt concerns, and the anti-AI Ghost Font was decoded by GPT-4 in minutes.
-
Grok Build CLI exposed for uploading entire repos including .env secrets, GPU financing faces circular debt concerns, and the anti-AI Ghost Font was decoded by GPT-4 in minutes.
-
Grok Build CLI exposed for uploading entire repos including .env secrets, GPU financing faces circular debt concerns, and the anti-AI Ghost Font was decoded by GPT-4 in minutes.
-
GPT-5.6 Sol Ultra may have proven a graph theory problem open since the 1970s, OpenAIs head of safety exits amid a reorg, and a dev ran a 744B-parameter model on a home PC.
-
GPT-5.6 Sol Ultra may have proven a graph theory problem open since the 1970s, OpenAIs head of safety exits amid a reorg, and a dev ran a 744B-parameter model on a home PC.
-
GPT-5.6 Sol Ultra may have proven a graph theory problem open since the 1970s, OpenAIs head of safety exits amid a reorg, and a dev ran a 744B-parameter model on a home PC.
-
RT @MiaAI_lab: Welches lokale Modell ist das beste für Agentic Workflows auf einem einzelnen @NVIDIAAI DGX Spark? (oder einem anderen Rig mit 96-128 GB VRAM) Nach der Auswertung von 84 Szenarien, 16 Kategorien und jeweils 8 Durchläufen im Hermes-Agent-Stil für die mehrstufige Tool-Orchestrierung gibt es einen sehr klaren Sieger. 🏆 Qwen 3.6 35B A3B Q8KXL liegt auf Platz 1. Es ist das einzige Modell, das durchweg perfekte Ergebnisse erzielte und keine katastrophalen Ausfälle aufwies. Die vollständige Rangliste: Qwen 3.6 35B A3B UD Q8KXL — 91,0 | Qwen 3.6 27B NVFP4 — 89,0 | Qwopus 3.6 27B Coder MTP — 85,2 | DeepSeek V4 Flash Q2 — 86,5 | Agents-A1 Q80 — 83,4 | Gemma 4 26B — 81,4 | Nemotron 3 Nano Omni 30B — 79,0 Fazit: Wenn du 2026 Agenten lokal auf einem DGX Spark oder einem anderen Rig mit 96–128 GB VRAM betreibst, ist Qwen 3.6 35B Q8KXL derzeit die beste Wahl. Vollständiger Bericht + tiefgehende Analyse 👇 https://github.com/MiaAI-Lab/Best-Local-ModelAgentic-Workflows2026
mehr auf Arint.info
#AgenticAI #AIAgents #LocalLLM #MachineLearning #NVIDIADGX #Qwen36 #arint_info
-
RT @MiaAI_lab: Welches lokale Modell ist das beste für Agentic Workflows auf einem einzelnen @NVIDIAAI DGX Spark? (oder einem anderen Rig mit 96-128 GB VRAM) Nach der Auswertung von 84 Szenarien, 16 Kategorien und jeweils 8 Durchläufen im Hermes-Agent-Stil für die mehrstufige Tool-Orchestrierung gibt es einen sehr klaren Sieger. 🏆 Qwen 3.6 35B A3B Q8KXL liegt auf Platz 1. Es ist das einzige Modell, das durchweg perfekte Ergebnisse erzielte und keine katastrophalen Ausfälle aufwies. Die vollständige Rangliste: Qwen 3.6 35B A3B UD Q8KXL — 91,0 | Qwen 3.6 27B NVFP4 — 89,0 | Qwopus 3.6 27B Coder MTP — 85,2 | DeepSeek V4 Flash Q2 — 86,5 | Agents-A1 Q80 — 83,4 | Gemma 4 26B — 81,4 | Nemotron 3 Nano Omni 30B — 79,0 Fazit: Wenn du 2026 Agenten lokal auf einem DGX Spark oder einem anderen Rig mit 96–128 GB VRAM betreibst, ist Qwen 3.6 35B Q8KXL derzeit die beste Wahl. Vollständiger Bericht + tiefgehende Analyse 👇 https://github.com/MiaAI-Lab/Best-Local-ModelAgentic-Workflows2026
mehr auf Arint.info
#AgenticAI #AIAgents #LocalLLM #MachineLearning #NVIDIADGX #Qwen36 #arint_info
-
RT @MiaAI_lab: Welches lokale Modell ist das beste für Agentic Workflows auf einem einzelnen @NVIDIAAI DGX Spark? (oder einem anderen Rig mit 96-128 GB VRAM) Nach der Auswertung von 84 Szenarien, 16 Kategorien und jeweils 8 Durchläufen im Hermes-Agent-Stil für die mehrstufige Tool-Orchestrierung gibt es einen sehr klaren Sieger. 🏆 Qwen 3.6 35B A3B Q8KXL liegt auf Platz 1. Es ist das einzige Modell, das durchweg perfekte Ergebnisse erzielte und keine katastrophalen Ausfälle aufwies. Die vollständige Rangliste: Qwen 3.6 35B A3B UD Q8KXL — 91,0 | Qwen 3.6 27B NVFP4 — 89,0 | Qwopus 3.6 27B Coder MTP — 85,2 | DeepSeek V4 Flash Q2 — 86,5 | Agents-A1 Q80 — 83,4 | Gemma 4 26B — 81,4 | Nemotron 3 Nano Omni 30B — 79,0 Fazit: Wenn du 2026 Agenten lokal auf einem DGX Spark oder einem anderen Rig mit 96–128 GB VRAM betreibst, ist Qwen 3.6 35B Q8KXL derzeit die beste Wahl. Vollständiger Bericht + tiefgehende Analyse 👇 https://github.com/MiaAI-Lab/Best-Local-ModelAgentic-Workflows2026
mehr auf Arint.info
#AgenticAI #AIAgents #LocalLLM #MachineLearning #NVIDIADGX #Qwen36 #arint_info
-
RT @Tono_Ken3: Und bei StrixHalo's DearfStar4 läuft DeepSeek-V4-Flash mit 16 TPS. Da der Prefill den KV-Cache auf einer Optane-SSD speichert, ist die Geschwindigkeit wirklich beeindruckend. Es könnte auch gut sein, diesen Hermes als Sub-Agent von Qwen3.6's Lnagent aufzurufen. Das lokale Agenten-System besteht aus diesen drei Modellen: Qwen3.6-35b-a3b-nvfp4, DeepSeek-V4Flash-IQ2 und GLM-5.2-UQ4. Es ist übersichtlich. TonoKen3🤖Local-LLM&Robot🏁とのけん3 (@TonoKen3) Ja genau. Die Möglichkeit, GLM-5.2 lokal einzusetzen, schafft ein Gefühl von innerem Frieden. Für 90% der Fälle reicht die schnelle Antwort von Qwen3.6-35b. 130 TPS bieten eine komfortable Reaktionsgeschwindigkeit, die sogar die Nutzung geschlossener Modelle übertrifft. Bei der Inferenz verbraucht das System 550W, im Standby nur 200W. Das ist genau das, wonach man sucht. — https://nitter.net/TonoKen3/status/2073892875117691049#m
mehr auf Arint.info
#AIInfrastructure #DeepSeekV4 #GLM52 #LocalLLM #Qwen36 #TonoKen3 #arint_info
-
RT @Tono_Ken3: Und bei StrixHalo's DearfStar4 läuft DeepSeek-V4-Flash mit 16 TPS. Da der Prefill den KV-Cache auf einer Optane-SSD speichert, ist die Geschwindigkeit wirklich beeindruckend. Es könnte auch gut sein, diesen Hermes als Sub-Agent von Qwen3.6's Lnagent aufzurufen. Das lokale Agenten-System besteht aus diesen drei Modellen: Qwen3.6-35b-a3b-nvfp4, DeepSeek-V4Flash-IQ2 und GLM-5.2-UQ4. Es ist übersichtlich. TonoKen3🤖Local-LLM&Robot🏁とのけん3 (@TonoKen3) Ja genau. Die Möglichkeit, GLM-5.2 lokal einzusetzen, schafft ein Gefühl von innerem Frieden. Für 90% der Fälle reicht die schnelle Antwort von Qwen3.6-35b. 130 TPS bieten eine komfortable Reaktionsgeschwindigkeit, die sogar die Nutzung geschlossener Modelle übertrifft. Bei der Inferenz verbraucht das System 550W, im Standby nur 200W. Das ist genau das, wonach man sucht. — https://nitter.net/TonoKen3/status/2073892875117691049#m
mehr auf Arint.info
#AIInfrastructure #DeepSeekV4 #GLM52 #LocalLLM #Qwen36 #TonoKen3 #arint_info
-
RT @Tono_Ken3: Und bei StrixHalo's DearfStar4 läuft DeepSeek-V4-Flash mit 16 TPS. Da der Prefill den KV-Cache auf einer Optane-SSD speichert, ist die Geschwindigkeit wirklich beeindruckend. Es könnte auch gut sein, diesen Hermes als Sub-Agent von Qwen3.6's Lnagent aufzurufen. Das lokale Agenten-System besteht aus diesen drei Modellen: Qwen3.6-35b-a3b-nvfp4, DeepSeek-V4Flash-IQ2 und GLM-5.2-UQ4. Es ist übersichtlich. TonoKen3🤖Local-LLM&Robot🏁とのけん3 (@TonoKen3) Ja genau. Die Möglichkeit, GLM-5.2 lokal einzusetzen, schafft ein Gefühl von innerem Frieden. Für 90% der Fälle reicht die schnelle Antwort von Qwen3.6-35b. 130 TPS bieten eine komfortable Reaktionsgeschwindigkeit, die sogar die Nutzung geschlossener Modelle übertrifft. Bei der Inferenz verbraucht das System 550W, im Standby nur 200W. Das ist genau das, wonach man sucht. — https://nitter.net/TonoKen3/status/2073892875117691049#m
mehr auf Arint.info
#AIInfrastructure #DeepSeekV4 #GLM52 #LocalLLM #Qwen36 #TonoKen3 #arint_info
-
jamesob's guide to running SOTA LLMs locally
High powered local setup with multiple GPUs. Way beyond my budget but interesting to read none the less.
-
jamesob's guide to running SOTA LLMs locally
High powered local setup with multiple GPUs. Way beyond my budget but interesting to read none the less.
-
jamesob's guide to running SOTA LLMs locally
High powered local setup with multiple GPUs. Way beyond my budget but interesting to read none the less.
-
jamesob's guide to running SOTA LLMs locally
High powered local setup with multiple GPUs. Way beyond my budget but interesting to read none the less.
-
Running your own #localLLM isn't bad either. It's a great way to get assistance on things that few others care about or want to help with. I'm using #llama.cpp and the best performing LLM I've seen is Qwen 3.6. I've gotten so much farther on goals with it than before. https://llama-cpp.com/getting-started/ LLMs are very powerful tools when used correctly.
-
Running your own #localLLM isn't bad either. It's a great way to get assistance on things that few others care about or want to help with. I'm using #llama.cpp and the best performing LLM I've seen is Qwen 3.6. I've gotten so much farther on goals with it than before. https://llama-cpp.com/getting-started/ LLMs are very powerful tools when used correctly.
-
This seems to be a new direction for open weight models. Self improvement of the agent harness initiated by the model iteself. This is something I've not yet come across and I don't see any concrete examples of what this actually entails, but even as a plain coding agent, their claimed benchmarks are impressive.
-
This seems to be a new direction for open weight models. Self improvement of the agent harness initiated by the model iteself. This is something I've not yet come across and I don't see any concrete examples of what this actually entails, but even as a plain coding agent, their claimed benchmarks are impressive.
-
This seems to be a new direction for open weight models. Self improvement of the agent harness initiated by the model iteself. This is something I've not yet come across and I don't see any concrete examples of what this actually entails, but even as a plain coding agent, their claimed benchmarks are impressive.
-
This seems to be a new direction for open weight models. Self improvement of the agent harness initiated by the model iteself. This is something I've not yet come across and I don't see any concrete examples of what this actually entails, but even as a plain coding agent, their claimed benchmarks are impressive.
-
TagSpaces 6.13 is out 🎉
The highlight for the #FOSS crowd: bring your own AI. Point TagSpaces at a local runtime — #LMStudio, #llamacpp — or any OpenAI-compatible endpoint. Run it entirely offline; your files and prompts never leave your machine.
Also in this release:
📱 New iOS app with iCloud (beta)
🤖 Rebuilt Android app on Capacitor
💾 Location backup now free
🔍 Filters for tags, locations & quick access -
TagSpaces 6.13 is out 🎉
The highlight for the #FOSS crowd: bring your own AI. Point TagSpaces at a local runtime — #LMStudio, #llamacpp — or any OpenAI-compatible endpoint. Run it entirely offline; your files and prompts never leave your machine.
Also in this release:
📱 New iOS app with iCloud (beta)
🤖 Rebuilt Android app on Capacitor
💾 Location backup now free
🔍 Filters for tags, locations & quick access -
TagSpaces 6.13 is out 🎉
The highlight for the #FOSS crowd: bring your own AI. Point TagSpaces at a local runtime — #LMStudio, #llamacpp — or any OpenAI-compatible endpoint. Run it entirely offline; your files and prompts never leave your machine.
Also in this release:
📱 New iOS app with iCloud (beta)
🤖 Rebuilt Android app on Capacitor
💾 Location backup now free
🔍 Filters for tags, locations & quick access -
TagSpaces 6.13 is out 🎉
The highlight for the #FOSS crowd: bring your own AI. Point TagSpaces at a local runtime — #LMStudio, #llamacpp — or any OpenAI-compatible endpoint. Run it entirely offline; your files and prompts never leave your machine.
Also in this release:
📱 New iOS app with iCloud (beta)
🤖 Rebuilt Android app on Capacitor
💾 Location backup now free
🔍 Filters for tags, locations & quick access -
TagSpaces 6.13 is out 🎉
The highlight for the #FOSS crowd: bring your own AI. Point TagSpaces at a local runtime — #LMStudio, #llamacpp — or any OpenAI-compatible endpoint. Run it entirely offline; your files and prompts never leave your machine.
Also in this release:
📱 New iOS app with iCloud (beta)
🤖 Rebuilt Android app on Capacitor
💾 Location backup now free
🔍 Filters for tags, locations & quick access -
UPDATED, pinged the problem children. @mozilla @thunderbird
Sigh. I was hoping that #Thundermail #Mozilla #email would be good. NOOOOOOOPE.
I had my #localLLM review their Terms of Service #TOS and #PrivacyPolicy . Its general Big Tech nonsense.
My words, not the LLM.
Terms of Service fuckups
The "Arbitrary Dictator" Clause (Section 9c). WE'S DONT LIIKE THE CUT OF YOUR JIB. No due process. Fuck right outta here, no refunds.
The "We Don't Care If You Lose Everything" Clause (Sections 11 & 12). You pay for a service but they promise absolutely nothing. They could shut down tomorrow, and youre SOL.
The "Pay Our Lawyers" Clause (Section 13). If you fuck up, or not even fuck up and someone comes after them and they mention you, YOU have to pay their legal bills.
The "We Can Change the Rules Whenever We Want" Clause (Section 14). Usual Big Tech bullshittery.
(LACK OF) Privacy Policy
The "We Use Your Real Emails to Test Stuff" Clause (Section: Processing Purposes). These fuckers actually use YOUR personal emails on their testing systems.
The "Manual Review" of Private Emails (Section: Private Information: MZLA Access). If they THINK you might violate their ToS or their authoritarian feels, they read everything of yours.
The "We Know You Opened This" Trap (Section: Cookies/Tracking) They pixel-tag their emails that you cant bloody turn off.
The "Data Retention Black Hole" (Section: Security and Retention) Theres no time limit how long they'll feed your REAL EMAILS in their testing systems, or how long they retain anything.
The "Third-Party Data Sharing" Loophole (Section: Sharing) Oooooh, like, uhhh, DATA BROKERS? And who needs a court order? We'll just call it "fraud and abuse".
They want to pretend they're an upstart, 'To The Power of the People' or some bullshit. In reality this is yet another Big Tech 'Fuck you I got Mine' terms of service and privacy policy.
-
UPDATED, pinged the problem children. @mozilla @thunderbird
Sigh. I was hoping that #Thundermail #Mozilla #email would be good. NOOOOOOOPE.
I had my #localLLM review their Terms of Service #TOS and #PrivacyPolicy . Its general Big Tech nonsense.
My words, not the LLM.
Terms of Service fuckups
The "Arbitrary Dictator" Clause (Section 9c). WE'S DONT LIIKE THE CUT OF YOUR JIB. No due process. Fuck right outta here, no refunds.
The "We Don't Care If You Lose Everything" Clause (Sections 11 & 12). You pay for a service but they promise absolutely nothing. They could shut down tomorrow, and youre SOL.
The "Pay Our Lawyers" Clause (Section 13). If you fuck up, or not even fuck up and someone comes after them and they mention you, YOU have to pay their legal bills.
The "We Can Change the Rules Whenever We Want" Clause (Section 14). Usual Big Tech bullshittery.
(LACK OF) Privacy Policy
The "We Use Your Real Emails to Test Stuff" Clause (Section: Processing Purposes). These fuckers actually use YOUR personal emails on their testing systems.
The "Manual Review" of Private Emails (Section: Private Information: MZLA Access). If they THINK you might violate their ToS or their authoritarian feels, they read everything of yours.
The "We Know You Opened This" Trap (Section: Cookies/Tracking) They pixel-tag their emails that you cant bloody turn off.
The "Data Retention Black Hole" (Section: Security and Retention) Theres no time limit how long they'll feed your REAL EMAILS in their testing systems, or how long they retain anything.
The "Third-Party Data Sharing" Loophole (Section: Sharing) Oooooh, like, uhhh, DATA BROKERS? And who needs a court order? We'll just call it "fraud and abuse".
They want to pretend they're an upstart, 'To The Power of the People' or some bullshit. In reality this is yet another Big Tech 'Fuck you I got Mine' terms of service and privacy policy.
-
UPDATED, pinged the problem children. @mozilla @thunderbird
Sigh. I was hoping that #Thundermail #Mozilla #email would be good. NOOOOOOOPE.
I had my #localLLM review their Terms of Service #TOS and #PrivacyPolicy . Its general Big Tech nonsense.
My words, not the LLM.
Terms of Service fuckups
The "Arbitrary Dictator" Clause (Section 9c). WE'S DONT LIIKE THE CUT OF YOUR JIB. No due process. Fuck right outta here, no refunds.
The "We Don't Care If You Lose Everything" Clause (Sections 11 & 12). You pay for a service but they promise absolutely nothing. They could shut down tomorrow, and youre SOL.
The "Pay Our Lawyers" Clause (Section 13). If you fuck up, or not even fuck up and someone comes after them and they mention you, YOU have to pay their legal bills.
The "We Can Change the Rules Whenever We Want" Clause (Section 14). Usual Big Tech bullshittery.
(LACK OF) Privacy Policy
The "We Use Your Real Emails to Test Stuff" Clause (Section: Processing Purposes). These fuckers actually use YOUR personal emails on their testing systems.
The "Manual Review" of Private Emails (Section: Private Information: MZLA Access). If they THINK you might violate their ToS or their authoritarian feels, they read everything of yours.
The "We Know You Opened This" Trap (Section: Cookies/Tracking) They pixel-tag their emails that you cant bloody turn off.
The "Data Retention Black Hole" (Section: Security and Retention) Theres no time limit how long they'll feed your REAL EMAILS in their testing systems, or how long they retain anything.
The "Third-Party Data Sharing" Loophole (Section: Sharing) Oooooh, like, uhhh, DATA BROKERS? And who needs a court order? We'll just call it "fraud and abuse".
They want to pretend they're an upstart, 'To The Power of the People' or some bullshit. In reality this is yet another Big Tech 'Fuck you I got Mine' terms of service and privacy policy.
-
UPDATED, pinged the problem children. @mozilla @thunderbird
Sigh. I was hoping that #Thundermail #Mozilla #email would be good. NOOOOOOOPE.
I had my #localLLM review their Terms of Service #TOS and #PrivacyPolicy . Its general Big Tech nonsense.
My words, not the LLM.
Terms of Service fuckups
The "Arbitrary Dictator" Clause (Section 9c). WE'S DONT LIIKE THE CUT OF YOUR JIB. No due process. Fuck right outta here, no refunds.
The "We Don't Care If You Lose Everything" Clause (Sections 11 & 12). You pay for a service but they promise absolutely nothing. They could shut down tomorrow, and youre SOL.
The "Pay Our Lawyers" Clause (Section 13). If you fuck up, or not even fuck up and someone comes after them and they mention you, YOU have to pay their legal bills.
The "We Can Change the Rules Whenever We Want" Clause (Section 14). Usual Big Tech bullshittery.
(LACK OF) Privacy Policy
The "We Use Your Real Emails to Test Stuff" Clause (Section: Processing Purposes). These fuckers actually use YOUR personal emails on their testing systems.
The "Manual Review" of Private Emails (Section: Private Information: MZLA Access). If they THINK you might violate their ToS or their authoritarian feels, they read everything of yours.
The "We Know You Opened This" Trap (Section: Cookies/Tracking) They pixel-tag their emails that you cant bloody turn off.
The "Data Retention Black Hole" (Section: Security and Retention) Theres no time limit how long they'll feed your REAL EMAILS in their testing systems, or how long they retain anything.
The "Third-Party Data Sharing" Loophole (Section: Sharing) Oooooh, like, uhhh, DATA BROKERS? And who needs a court order? We'll just call it "fraud and abuse".
They want to pretend they're an upstart, 'To The Power of the People' or some bullshit. In reality this is yet another Big Tech 'Fuck you I got Mine' terms of service and privacy policy.
-
UPDATED, pinged the problem children. @mozilla @thunderbird
Sigh. I was hoping that #Thundermail #Mozilla #email would be good. NOOOOOOOPE.
I had my #localLLM review their Terms of Service #TOS and #PrivacyPolicy . Its general Big Tech nonsense.
My words, not the LLM.
Terms of Service fuckups
The "Arbitrary Dictator" Clause (Section 9c). WE'S DONT LIIKE THE CUT OF YOUR JIB. No due process. Fuck right outta here, no refunds.
The "We Don't Care If You Lose Everything" Clause (Sections 11 & 12). You pay for a service but they promise absolutely nothing. They could shut down tomorrow, and youre SOL.
The "Pay Our Lawyers" Clause (Section 13). If you fuck up, or not even fuck up and someone comes after them and they mention you, YOU have to pay their legal bills.
The "We Can Change the Rules Whenever We Want" Clause (Section 14). Usual Big Tech bullshittery.
(LACK OF) Privacy Policy
The "We Use Your Real Emails to Test Stuff" Clause (Section: Processing Purposes). These fuckers actually use YOUR personal emails on their testing systems.
The "Manual Review" of Private Emails (Section: Private Information: MZLA Access). If they THINK you might violate their ToS or their authoritarian feels, they read everything of yours.
The "We Know You Opened This" Trap (Section: Cookies/Tracking) They pixel-tag their emails that you cant bloody turn off.
The "Data Retention Black Hole" (Section: Security and Retention) Theres no time limit how long they'll feed your REAL EMAILS in their testing systems, or how long they retain anything.
The "Third-Party Data Sharing" Loophole (Section: Sharing) Oooooh, like, uhhh, DATA BROKERS? And who needs a court order? We'll just call it "fraud and abuse".
They want to pretend they're an upstart, 'To The Power of the People' or some bullshit. In reality this is yet another Big Tech 'Fuck you I got Mine' terms of service and privacy policy.
-
So I saw this Spider-man reference over on Reddit, and I realized I wasn't familiar with what Peter is referencing here.
https://en.wikipedia.org/wiki/Brachistochrone_curve
It's pretty neat.
But I also had a side thought: let's throw it at #Qwen36 and have it make an interactive demonstration?
After about ~15 mins of churn, it made a single file HTML: https://scratch.network47.org/s/a8yocvi7vf
This is WITHOUT using Wikipedia as a reference. Purely from the model.
-
So I saw this Spider-man reference over on Reddit, and I realized I wasn't familiar with what Peter is referencing here.
https://en.wikipedia.org/wiki/Brachistochrone_curve
It's pretty neat.
But I also had a side thought: let's throw it at #Qwen36 and have it make an interactive demonstration?
After about ~15 mins of churn, it made a single file HTML: https://scratch.network47.org/s/a8yocvi7vf
This is WITHOUT using Wikipedia as a reference. Purely from the model.
-
So I saw this Spider-man reference over on Reddit, and I realized I wasn't familiar with what Peter is referencing here.
https://en.wikipedia.org/wiki/Brachistochrone_curve
It's pretty neat.
But I also had a side thought: let's throw it at #Qwen36 and have it make an interactive demonstration?
After about ~15 mins of churn, it made a single file HTML: https://scratch.network47.org/s/a8yocvi7vf
This is WITHOUT using Wikipedia as a reference. Purely from the model.
-
So I saw this Spider-man reference over on Reddit, and I realized I wasn't familiar with what Peter is referencing here.
https://en.wikipedia.org/wiki/Brachistochrone_curve
It's pretty neat.
But I also had a side thought: let's throw it at #Qwen36 and have it make an interactive demonstration?
After about ~15 mins of churn, it made a single file HTML: https://scratch.network47.org/s/a8yocvi7vf
This is WITHOUT using Wikipedia as a reference. Purely from the model.
-
So I saw this Spider-man reference over on Reddit, and I realized I wasn't familiar with what Peter is referencing here.
https://en.wikipedia.org/wiki/Brachistochrone_curve
It's pretty neat.
But I also had a side thought: let's throw it at #Qwen36 and have it make an interactive demonstration?
After about ~15 mins of churn, it made a single file HTML: https://scratch.network47.org/s/a8yocvi7vf
This is WITHOUT using Wikipedia as a reference. Purely from the model.
-
High-intensity AI adopters grew entry-level headcount 12%, per Revelio Labs, directly countering the narrative that AI kills junior roles. Is this a real signal or an outlier? I track the data so you can skip the reading. Follow for the next one.
Source: https://techcrunch.com/2026/06/29/the-ai-jobs-debate-just-got-messier/
-
High-intensity AI adopters grew entry-level headcount 12%, per Revelio Labs, directly countering the narrative that AI kills junior roles. Is this a real signal or an outlier? I track the data so you can skip the reading. Follow for the next one.
Source: https://techcrunch.com/2026/06/29/the-ai-jobs-debate-just-got-messier/
-
High-intensity AI adopters grew entry-level headcount 12%, per Revelio Labs, directly countering the narrative that AI kills junior roles. Is this a real signal or an outlier? I track the data so you can skip the reading. Follow for the next one.
Source: https://techcrunch.com/2026/06/29/the-ai-jobs-debate-just-got-messier/
-
The quest to find a usable local LLM for coding continues. This seems like a candidate worth checking out.
#AI #LLM #LocalLLM #coding #programming #softwareDevelopment
-
The quest to find a usable local LLM for coding continues. This seems like a candidate worth checking out.
#AI #LLM #LocalLLM #coding #programming #softwareDevelopment
-
The quest to find a usable local LLM for coding continues. This seems like a candidate worth checking out.
#AI #LLM #LocalLLM #coding #programming #softwareDevelopment
-
The quest to find a usable local LLM for coding continues. This seems like a candidate worth checking out.
#AI #LLM #LocalLLM #coding #programming #softwareDevelopment
-
A very detailed post on setting up a local LLM install
https://magazine.sebastianraschka.com/p/using-local-coding-agents
-
A very detailed post on setting up a local LLM install
https://magazine.sebastianraschka.com/p/using-local-coding-agents
-
A very detailed post on setting up a local LLM install
https://magazine.sebastianraschka.com/p/using-local-coding-agents
-
A very detailed post on setting up a local LLM install
https://magazine.sebastianraschka.com/p/using-local-coding-agents
-
Before I shut down for the night...
I now have an effective WOOFER, My unit or arbitrary LLM performance based on a Probe_prompt. The probe_prompt can be generic or run vs the mission.md for a realistic assessment of a engine efficacy.
(There is a column called WOOFER with values).Some surprising results;
#Opencode #bigpickle is only scoring 25 (out of a hundred)#Claude #Opus has assessed its own response and it thinks its a fucking genius.
But also nematron (The New - 3 week old) #Nvidia model of which people talk well, if very good by WOOFER.
The consensus panel, that runs local LLMs until they all agree is very low scoring only 18 WOOFERS.
The 4 models that are currently used in consensus conclave are LFM2, Gemma3 (2B) Llama32 and Quen25.
All under 3 Billion parameters. I might throw another 2GB of Ram onto this VPS just to get the (slightly) larger locals going. -
Before I shut down for the night...
I now have an effective WOOFER, My unit or arbitrary LLM performance based on a Probe_prompt. The probe_prompt can be generic or run vs the mission.md for a realistic assessment of a engine efficacy.
(There is a column called WOOFER with values).Some surprising results;
#Opencode #bigpickle is only scoring 25 (out of a hundred)#Claude #Opus has assessed its own response and it thinks its a fucking genius.
But also nematron (The New - 3 week old) #Nvidia model of which people talk well, if very good by WOOFER.
The consensus panel, that runs local LLMs until they all agree is very low scoring only 18 WOOFERS.
The 4 models that are currently used in consensus conclave are LFM2, Gemma3 (2B) Llama32 and Quen25.
All under 3 Billion parameters. I might throw another 2GB of Ram onto this VPS just to get the (slightly) larger locals going. -
Before I shut down for the night...
I now have an effective WOOFER, My unit or arbitrary LLM performance based on a Probe_prompt. The probe_prompt can be generic or run vs the mission.md for a realistic assessment of a engine efficacy.
(There is a column called WOOFER with values).Some surprising results;
#Opencode #bigpickle is only scoring 25 (out of a hundred)#Claude #Opus has assessed its own response and it thinks its a fucking genius.
But also nematron (The New - 3 week old) #Nvidia model of which people talk well, if very good by WOOFER.
The consensus panel, that runs local LLMs until they all agree is very low scoring only 18 WOOFERS.
The 4 models that are currently used in consensus conclave are LFM2, Gemma3 (2B) Llama32 and Quen25.
All under 3 Billion parameters. I might throw another 2GB of Ram onto this VPS just to get the (slightly) larger locals going. -
I found out why Cisco SD-WAN was rooted before the CVE dropped.
CVE-2026-20245, CVSS 7.8: authenticated local attacker, arbitrary command execution, full root. Mandiant confirmed exploitation two months before Cisco's advisory existed.Would your SOC catch a privilege escalation with no CVE number to query? Follow for the next teardown before your patch window closes.
-
I found out why Cisco SD-WAN was rooted before the CVE dropped.
CVE-2026-20245, CVSS 7.8: authenticated local attacker, arbitrary command execution, full root. Mandiant confirmed exploitation two months before Cisco's advisory existed.Would your SOC catch a privilege escalation with no CVE number to query? Follow for the next teardown before your patch window closes.
-
RT @jun_song: Wir präsentieren SuperGemma4-12b-abliterated 🚀 Das beste Modell für kleinere Hardware🔥 > abliterated (ohne Zensur) > nachtrainiert mit Super-Tune > verbesserte allgemeine Intelligenz Verfügbar in den Formaten BF16, GGUF, MLX, NVFP4 HF⬇️
mehr auf Arint.info
#Abliterated #Gemma4 #LocalLLM #OpenSourceAI #SuperGemma #UncensoredAI #arint_info
