Lucas Catão de Moraes (@LucasDolutech) header
Lucas Catão de Moraes (@LucasDolutech) profile picture

Founder @ CEO - DolutechAI | Researcher and Specialist in Cybersecurity and Artificial Intelligence | RL - SFT - DPO

Portugal
Joined August 2024
Photo by Lucas Catão de Moraes (@LucasDolutech)Photo by Lucas Catão de Moraes (@LucasDolutech)Photo by Lucas Catão de Moraes (@LucasDolutech)Photo by Lucas Catão de Moraes (@LucasDolutech)Photo by Lucas Catão de Moraes (@LucasDolutech)Photo by Lucas Catão de Moraes (@LucasDolutech)Photo by Lucas Catão de Moraes (@LucasDolutech)Photo by Lucas Catão de Moraes (@LucasDolutech)Photo by Lucas Catão de Moraes (@LucasDolutech)Photo by Lucas Catão de Moraes (@LucasDolutech)
I saw something quite curious today! An acquaintance of mine fell victim to a cyberattack, and as an ethical hacker, I decided to take a closer look at what happened. What surprised me was that the attackers were using OpenCode, without any actual hacking tools (lol). Their entire approach relied solely on "webfetch", "request_http", and "playwright". Brute force attacks, persistence mechanisms, and automation, all carried out through basic HTTP requests and browser interactions. It's fascinating how even simple tools can be misused to carry out cyberattacks.
2
215
Lucas Catão de Moraes retweeted
Day 13/ Le limit reset
1,634
632
823
16,398
1,794,858
Replying to @LucasDolutech
Ótimos insights! Sensacional a sua contribuição
1
2
188
Lucas Catão de Moraes retweeted
"A large scale distillation attack". Yeah, sure. Come on guys. This is a security breach. Call things by their names. @thdxr what are you doing? You are good guys. Dont do this corporate bullshit. We all understand. OpenCode had a breach. Stored OpenAI, Anthropic and Google keys of 1.4% of Console users were exposed, and people are posting drained balances. $400 overnight, $1,700 for another guy. This is not distillation. Someone got into your system and took peoples keys. But "distillation" is the scary word of the year, so now every bad thing gets that label. Stolen keys, bans, export controls, all of it. My local models dont have an API key. There is nothing on my rig to drain.
We had a security breach last night related to a large scale distillation attack that impacted 1.4% of OpenCode Console users We have notified them all this morning via email. If you did not get an email you were not impacted, it was a subset of users who had external providers linked The incident is ongoing so we cannot share more details at this point but we are working with Anthropic, OpenAI, and Google to remediate
44
20
5
464
47,022
Lucas Catão de Moraes retweeted
so many of you dumbasses doubted us never do that
Operation Cheepseek Phase 3: $80 of usage for DeepSeek V4.1 Flash
173
44
8
3,169
120,312
Lucas Catão de Moraes retweeted
Replying to @LucasDolutech
Bouncing back after a 1M venture is no joke! A massive respect for the resilience. Building RecT Solutions over here in the tech space. Would love to connect and talk startups/tech!
1
1
22
Lucas Catão de Moraes retweeted
Meet Mistral Large 4, aka Le Chonk. • 1T parameters, natively multimodal. 49B active. It is the best open weights model from US or Europe on aggregated benchmarks. • State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding. • Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure. • Available to all via API today. Working with cybersecurity partners privately. Open weights release end of October.
🤖 Made with AI
2,034
4,563
2,812
40,613
5,761,982
Lucas Catão de Moraes retweeted
Cognition is the first customer running @NVIDIA Vera Rubin, powered by @CoreWeave! Hopper in 2022. Blackwell in 2024. Now Vera Rubin: On SWE-2 inference, the new chips deliver about 4.8x more token throughput than GB200 at the same decode speed. More compute, better agents!
50
58
31
682
157,415
Lucas Catão de Moraes retweeted
As a cybersecurity and AI specialist, I can say that GLM 5.3 is more beneficial to cybersecurity than harmful! Today, with Habilateralidade, we can remove the guardrails of many models! And GLM 5.3 is a great help in combating cyberattacks! The problem is that people lacking knowledge don't know this! While closed models from Anthropic and OpenAI are only accessible to "special" partners.
3
6
1
104
3,242
Lucas Catão de Moraes retweeted
Replying to @DulcesYupis @OpenAI
I guess the model sucked in benchmarks.
1
2
23
1,420
Lucas Catão de Moraes retweeted
Replying to @claudeai
how about a friendly traditional weekly reset
11
2
349
14,529
Lucas Catão de Moraes retweeted
Replying to @cognition
@cognition One thing I can say for sure! Devin's Cloud environment is the best available on the market! It's the only one that has macOS, Windows, and Ubuntu! I'm a Linux user and thanks to Cloud Agents I can compile my .dmg connectors for my clients perfectly.
1
1
49
Lucas Catão de Moraes retweeted
Founder sales trap: every deal starts with the founder explaining the product from scratch.
2
2
2
35
Lucas Catão de Moraes retweeted
A founder can spend 60 hours in meetings and leave the company exactly where it was. Busy is a lagging metric. Measure the week by decisions made and bottlenecks removed. Check whether customers got a better result.
2
2
17
Lucas Catão de Moraes retweeted
this is taco bell's new ai model
Space Bunny (stealth model) is free for the next week - 1M Context - Multi-modal - Zero Data Retention
77
24
5
2,054
116,503
Lucas Catão de Moraes retweeted
The race for AGI Script: Sherpa by Pocket FM Video: Seedance 2.5
🤖 Made with AI
852
3,466
1,046
21,768
4,141,731
Lucas Catão de Moraes retweeted
Tip of the day! Anthropic's subscription is much more generous than OpenAI's. I'm talking about 5 to 6 times more! It's unbelievable how GPT 6 Sol exhausts its weekly subscription on the plan and can barely finish fixing a module...
1
1
2
89
Tip of the day! Anthropic's subscription is much more generous than OpenAI's. I'm talking about 5 to 6 times more! It's unbelievable how GPT 6 Sol exhausts its weekly subscription on the plan and can barely finish fixing a module...
1
1
2
89
GPT6 Sol exhausted the Pro 5x plan in 1 security check implementation task... while Opus 5.5 manages to do 5x more in an equivalent plan on Anthropic.
1
97