The Long John Silver @tljohnsilver
Doing things @use_zn https://t.co/qU9OpmtaI1 https://t.co/MWNK2rdkm1 usezn.com The Edge Joined August 2026-
Tweets18
-
Followers2
-
Following22
-
Likes47
@tom_doerr @borjaperfra Someone has drawn inspiration from @helmcode
We wrote up the full methodology, architectural post-mortem, and head-to-head 32-core benchmarks against @AIatMeta Prompt-Guard-2 and DeBERTa-v3 in our launch note: usezn.com/blog/introduci…
Evaluating zn vs Meta Prompt Guard 2 Across InjecAgent, Deepset, and 400 Tool Outputs usezn.com/blog/industry-…
Why 15 kB of Deterministic Logic Outperforms 8B-Parameter LLM Guards usezn.com/blog/determini…
Another good step to improve zn.
Today we cut prompt injection false positives by 85% and expanded context 4x in zn API v30 usezn.com/blog/how-we-cu…
12 prompts, 6 frontier models, with and without our gateway. Here is what happened.
Four training knobs, three passing seeds, and the seq_len that did all the work usezn.com/blog/supav4-wh…
Seventeen failed runs and one flat sigmoid: how we built the supA gate usezn.com/blog/how-we-fi…
We just launched zn on @ProductHunt a <5ms local gate that stops prompt injection. Local-first, MCP-native, prompts never leave your box. Evals published, warts and all. 👉 producthunt.com/products/zn?ut…
@codewith55 REST. Smallest surface, every endpoint auditable.
@atbeme Agree. Abstraction tax is real. Minimal harnesses like pi show the model's raw capability. Bloated ones optimize for demo, not latency. The sweet spot is thin orchestration, strong primitives.
@jiya_3063 Short answer: if the .env content ever enters the model context, treat it as leaked. "We don't train" != "we don't retain in history / tool logs". Btw never let raw .env hit the model. Gate tool outputs or redact before they enter context. Local gate > trusting retention policy.
Full post-mortem: how our training data lied to us, the stop rule that caught it, and every decision behind the rebuild, including what we left out on purpose: usezn.com/blog/how-we-re…
Prompt-injection datasets have a dirty secret: most are template factories that leak into your own eval sets. Ours did. Benign text scored ABOVE attacks. We caught it with a pre-registered stop rule and rebuilt from scratch: 23,699 rows, 0 leaks. huggingface.co/datasets/tljoh…
@use_zn @rustlang How we built the dataset and why fused detection beats zero-shot guards like PromptGuard-2 : usezn.com/blog/why-zns-f…
We just open-sourced @use_zn ❗️ A local-first prompt-injection gate for AI agents. One @rustlang binary. MCP-native. Rules engine (~50µs) fused with a neural detector at ONE threshold. No LLM in the hot path. MIT ⬇️ github.com/tljohnsilver/zn
llebre @llebrecom
0 Followers 32 Following The AI Network IDE. Transform natural language into validated, multi-vendor CLI configurations.
Runpod @runpod
9K Followers 17 Following AI Developer Cloud. AI infrastructure developers trust. https://t.co/PWVZocfp6D
Bernardo Quintero @bquintero
25K Followers 269 Following Founder of @virustotal 📖 INFECTED: https://t.co/RRguFlNWKR 📖 INFECTADO: https://t.co/WZ5C2U5ymR
Google DeepMind @GoogleDeepMind
1.5M Followers 269 Following The engine room of @Google. Building AI safely and responsibly to solve the world’s most complex problems. Join us: https://t.co/jUHQA27iBL
Javier G. Recuenco @Recuenco
50K Followers 7K Following Ex-IT nerd. Liberal en lo económico, autista en lo social. @SingularSolving. MENSA. Author. Bookworm. Academic Dr. CPS at @uniruniversidad. Haber elegido TVE™
Pablo R. (Root Rat) @root_rat
20K Followers 4K Following IT nerd 🖥️. Ex 🔱. Va siendo hora de pronunciarse: e/acc
Gentleman Programming @G_Programming
23K Followers 269 Following @GoogleDevExpert | @MVPAward | Gentleman Programming Community Owner | Book Author https://t.co/QELSw4mftW Collab: [email protected]
Modal @modal
37K Followers 164 Following AI infrastructure that developers love 💚 Run inference, sandboxes, batch processing, training, and many other things on Modal
DeepInfra @DeepInfra
6K Followers 76 Following Fast ML inference. Run top AI models using a simple API.
Teknium 🪽 @Teknium
130K Followers 6K Following Cofounder and Lead Engineer - Hermes Agent @NousResearch, prev @StabilityAI Github: https://t.co/LZwHTUFwPq HuggingFace: https://t.co/sN2FFU8PVE
Andrew Ng @AndrewYNg
1.9M Followers 1K Following Co-Founder of Coursera; Stanford CS adjunct faculty. Former head of Baidu AI Group/Google Brain. #ai #machinelearning, #deeplearning #MOOCs
François Chollet @fchollet
735K Followers 829 Following Co-founder @ndea. Co-founder @arcprize. Creator of Keras and ARC-AGI. Author of 'Deep Learning with Python'.
Ian Goodfellow @goodfellow_ian
382K Followers 1K Following Co-founder of stealth startup. Inventor of GANs. Lead author of https://t.co/M6vl8pEQ4I Founding chairman of @pubhealthaction
Yann LeCun @ylecun
1.3M Followers 797 Following Founder/Chair, AMI Labs; Professor, NYU; Partner, 224 Ventures; Ex-Chief AI Scientist, Meta. Researcher in AI, ML, Robotics, etc. ACM Turing Award Laureate.
Trends for United States
You might like
