Alex Teo (@AlexGTeo) header
Alex Teo (@AlexGTeo) profile picture

AI builder helping people learn and feel more confident about the future. Building and evaluating my own agents. Exploring EEG & BCI. 🇸🇬+SF+SZ

Joined May 2009
Photo by Alex Teo (@AlexGTeo)Photo by Alex Teo (@AlexGTeo)Photo by Alex Teo (@AlexGTeo)Photo by Alex Teo (@AlexGTeo)Photo by Alex Teo (@AlexGTeo)Photo by Alex Teo (@AlexGTeo)Photo by Alex Teo (@AlexGTeo)Photo by Alex Teo (@AlexGTeo)Photo by Alex Teo (@AlexGTeo)Photo by Alex Teo (@AlexGTeo)
1/4 Building a DIY EEG wasn't on my 2026 bingo card, yet somehow here we are. 🧠 I've been exploring neuroscience and brain-computer interfaces (BCIs), particularly how they might help people learn and retain information. Figured I'd get my hands dirty and build my own 8-channel EEG.
2
1
4
168
Today we’re releasing FrontierBreach, a new standard for attack success rate on agentic workflows in the enterprise. Existing benchmarks systematically underrepresent the security risks facing modern AI agents. FrontierBreach reflects attack scenarios on realistic enterprise workflows, while using sophisticated attack chains that result in cyber consequences (ransomware, data exfiltration, phishing, etc).
167
171
90
1,721
1,216,742
I'm hopeful of the possibility too! To borrow an analogy: • Invasive BCI (eg Neuralink) is like attending a rock concert inside the stadium. You get the signals more directly. • Meanwhile, non-invasive BCI is standing in the stadium carpark trying to hear the what's going on. The signal is filtered through walls (i.e. your skull) and might pick up all kinds of noise along the way.
Replying to @alljoined
The signal we're using in this demo is 10x weaker than what Neuralink is using today. Brain activity is very hard to read non-invasively because neurons emit very small voltages and by the time it propagates outside the skull it’s almost all noise. But we believe that deep learning can extract signal from this noise and our deep learning models make sense of the smallest fluctuations to create a representation that computers can use.
56
Most tech companies over the last 20 years: • "It's always Day 1" Most companies over the last 20 days: • "It's Zero Day everyday"
This is a big one. We found malicious JavaScript being served through Optimizely to some of the world's largest websites. 178 organisations affected, with everything from crypto theft and card skimming to iOS exploits. Zoom, BBC, CNN, HSBC, Salesforce, GitLab, KLM, IKEA, Equifax, Ledger. The list goes on... Full list in the post. blog.report-uri.com/worlds-l…
1
163
This sounds obvious on hindsight, but it's a bad idea to: • Set up your database/backend in Japan; • Have the frontend in us-central; • Do a user demo in Singapore. Thank goodness Claude could help with the db migration (without deleting my data 😝)
1
1
86
#Claude Max with Fable 5.2 = "Code-Vember"
1
67
1/2 “Use the Force…” Still can’t reliably move a cursor with my brainwaves 🙂 I spent a week recording my own EEG with the 8-channel device I built. In the session shown here, the alpha peak around 9.3 Hz was stronger with my eyes closed, as expected. #EEG
3
4
135
2/2 It’s become a bit of a daily routine, almost like meditating. I sit still and suddenly notice the toaster dinging, a bus reversing, or a bulbul chirping. I wrote up the first week of recordings here 👇 alexteo.com/blog/daily-eeg-s…
31
That's a first. (Thank goodness I'm not Chinese superstitious.) P.S. maybe it was meant for MS 365 🫢
1
52
This feels like the prelude to a, "Senator, I'm Singaporean" moment.
1
95
Inspired by Black October's security incidents
For your playlist the next time there's a security incident (which should be like.. Next Tuesday?) youtu.be/J8U712cpsTQ
1
60
3/4 Got Grok Bot to help with the shopping list, liaise with the PCB manufacturer, and even create personalised IKEA-style assembly instructions. Software is a relative stroll in the park compared to hardware + software. 😅 Still testing, so no reliable EEG data just yet.
2
3
36
Darn right Grok Bot, darn right.
57
If space is the final frontier, then I guess we have a new class of 'frontier' models
A ‘hello world’ milestone for Project Suncatcher 🚀 After launching our prototype satellite a few weeks ago, we’ve now successfully sent a command up to orbit, booted up the TPU, ran the prompt, and received our first complete response back on earth - a haiku!
52
RIP Professor
In lieu of flowers, memorial contributions may be made to: Johns Hopkins Alzheimer’s Disease Research Center Fund    alzresearch.org/giving/ Alzheimer’s Foundation of America Caregiver Resources     alzfdn.org/donate/ Photo Credit: Tommy Garcia
31
3 weeks ago, I : - Cloned a public repo; - Prompted an agent to make some edits; - Pushed those updates manually. But 3 weeks on, the repo exists and I have no idea which model/harness I used. Claude, Codex and Cursor all say there have no past records of working on the repo. Who's lying? Or did I make those edits manually? Or is there another model/agent at work here? Perplexing..
1
1
83
That feeling when your promising ML experiment survives a second random seed 🥳 10x the training exposure for ~4x the retrieval accuracy. More data exposure, more accuracy. Who would've thought? 🤷‍♂️ #eeg #bci #neurotech #machinelearning
1
56
Putting those college classes on Art History to some use.
Is the fate of all art to become commoditized slop? And is AI art just speed-running the process? We take you on a trip through 500 years of art history with this one: youtu.be/GOoKe1ZHdz0
1
37