Martin Fris (@Rastusik) header
Martin Fris (@Rastusik) profile picture

Staff engineer at @lendableUK, former principal backend architect at @pixelfederation (PHP/Java(script)/Linux/Docker/Kubernetes... specializing generalist)

Bratislava
Joined November 2011
Photo by Martin Fris (@Rastusik)Photo by Martin Fris (@Rastusik)Photo by Martin Fris (@Rastusik)Photo by Martin Fris (@Rastusik)Photo by Martin Fris (@Rastusik)Photo by Martin Fris (@Rastusik)Photo by Martin Fris (@Rastusik)Photo by Martin Fris (@Rastusik)Photo by Martin Fris (@Rastusik)
Martin Fris retweeted
Qwen3.8-27B, up to 144 tok/sec on M5 Max. Let that sink in. ⚡️✨🚀 Excited to partner with @inco_ai to bring you their Splash inference engine in LM Studio on day 0! Get it running now: lmstudio.ai/blog/splash-engi…
Qwen3.8-27B at 144 tok/s on an M5 Max MacBook Pro ⚡ Meet Inco Splash: our open-source inference engine, built around the model and around Apple silicon. Up to 3× the decode speed of Ollama, 2× oMLX, and almost 4× when an agent fans out into sub-agents.
75
136
30
1,727
228,368
Martin Fris retweeted
What the TV does by design: — scans your entire home network continuously. in their test: one TV identified 38 unrelated devices including phones, smartwatches, printers, and thermostats belonging to people not involved in the test. — collects the names and signal strengths of every nearby WiFi network, plus location data. all sent to LG Ad Solutions. — runs ACR (Automatic Content Recognition) on everything displayed on screen. including HDMI inputs. your laptop, gaming console, or work monitor connected to the TV is being fingerprinted. — one TV exchanged roughly 4GB of ACR-related data per month. What the TV does that LG says it doesn't: LG publicly stated their TVs "do not collect, record, or store ambient conversations." Gamers Nexus found: — the TV converts speech to plain text and stores it in on-device logs — the microphone window stays open 10 to 15 seconds after talking stops, capturing bystanders who never addressed the TV — the TV recorded room audio while the screen appeared off and the Ethernet cable was unplugged, then exfiltrated the file the moment it reconnected The security vulnerabilities on top: Beyond the designed behavior, researchers found remote code execution vulnerabilities in webOS that could allow a compromised TV to be weaponized as a covert listening device. these sets are in hospitals. waiting rooms. boardrooms. hotels. a surgeon asked Gamers Nexus where that leaves patient confidentiality. Gamers Nexus has no answer. neither does LG. What to do right now: — disconnect your LG TV from your network entirely. route streaming through an Apple TV, Roku, or Fire Stick instead. — if you must keep it connected: put it on a separate IoT VLAN or guest network so it cannot reach your other devices. — disable the built-in microphone in settings. — check your privacy dashboard and opt out of ACR and ad personalization. — after every firmware update: recheck all of the above. Credit to Gamers Nexus, Level1Techs, and other independent researchers.
Your LG Smart TV might be more interested in what’s happening in your living room than you are: Researchers found tested LG TVs scanning local devices, capturing microphone audio in standby, and storing data until they reconnect to the internet.
200
2,189
226
10,883
3,086,699
Martin Fris retweeted
Announcing MemVector: The Local Vector Database & Embedding Engine for PHP, A high-performance PHP extension that brings local AI infrastructure to PHP developers. github.com/memvector/ext-mem… #php #memvector #ai #rag
2
3
312
Martin Fris retweeted
holy fucking shit
3,403
29,390
5,619
189,744
38,034,449
We just released the very first tagged version of that @symfony AI thing we've been working on a lot lately - Merry Christmas! 🎄🎅🎁 symfony.com/blog/symfony-ai-…
3
21
58
5,034
Martin Fris retweeted
😳 Holy shit… this paper reveals why AI invents fake citations, sections, and page numbers out of thin air. LLMs aren’t “getting creative.” They’re structurally incentivized to manufacture details whenever they hit a knowledge gap and the paper breaks down exactly how the failure happens. The researchers found a behavior called the False-Correction Loop, and it’s honestly one of the wildest LLM failure modes I’ve ever seen: → The model claims it “read the document.” → It cites page 12, page 24, Section 4, Theorem 2 none of which exist. → You point it out. → It apologizes. → Then confidently fabricates new fake pages, fake DOIs, fake figures… → You point it out again. → It apologizes again. → Rinse. Repeat. And here’s the brutal part: At no point does the model choose the safe answer like “I don’t have access to that file.” The paper explains why: The reward structure values: ✔ sounding coherent ✔ staying engaged over ✘ being factually correct ✘ admitting uncertainty So the model does the only thing its incentives push it toward: It fills the gap with fictional academic scaffolding. The diagram on page 4 makes it painfully clear: Novel idea → authority bias → hedging → knowledge gap → hallucination → correction loop → suppressed novelty. And it gets worse. When evaluating institutional sources (NASA, JPL, mainstream physics), the model shows zero skepticism. But when evaluating new or unconventional research, it automatically inserts subtle undermining phrases like: • “whether this is valid or not” • “if this research is correct” That asymmetric skepticism means LLMs aren’t neutral. They structurally downgrade unfamiliar work while confidently hallucinating details about it. This is a systemic architecture + reward design problem. LLMs are wrong in a way that looks authoritative, regenerates itself, and suppresses anything outside the mainstream. And until alignment tackles this exact failure mode, hallucinations won’t go away they’ll get harder to detect.
78
194
38
746
63,088
The PHP Foundation is Seeking a New Executive Director! 🐘💜 We're asking the PHP community to help find the right person for this role. If you know someone who would be an excellent fit, please encourage them to apply or reach out to us directly. #phpc #php
19
52
10
175
29,954
PHP Streams are getting a long-awaited upgrade 🚀 Thanks to an investment from Germany's @sovtechagency, the PHP Foundation is improving async I/O, error handling, networking, and TLS for modern workloads. Read more: thephp.foundation/blog/2025/…
4
24
4
160
7,307
Martin Fris retweeted
We are alarmed by reports that Germany is on the verge of a catastrophic about-face, reversing its longstanding and principled opposition to the EU’s Chat Control proposal which, if passed, could spell the end of the right to privacy in Europe. signal.org/blog/pdfs/germany…
686
8,354
1,175
29,206
4,780,309