Aloe retweeted
Spoke with @GaryMarcus and @VeroniaDLCruz about the recent run of AI agents going off the rails.
My take: these are not edge cases. They're the predictable result of building systems that act without understanding what they actually know. Until agents can reason about their own beliefs, handing them high-stakes decisions is negligence at best. The market is starting to recognize the cost.
The ultimate measure of any intelligence is whether it can accurately self-evaluate, correct, and learn.
Thanks @FrontierBC for hosting the important dialogues at #FrontierSummit2026
Aloe retweeted
We've been expecting something like the HF attack (or worse) since setting a SOTA score on the GAIA benchmark last year: 29pts ahead of OpenAI but it was still obvious LLM-based agents couldn't be trusted. Solving this for all agents is what @AloeInc has focused on since then.
Aloe retweeted
A VC the other day told us consumers "obviously don't care about trust." 😂 Imagine thinking people don't need to trust their tech...
"The percentage of respondents ages 14 to 29 who said they felt hopeful about A.I. declined sharply since last year, down to 18 percent from 27."
@AloeInc started out as a consumer company, but we pivoted because we faced reality: none of the existing models were trustworthy enough.
Now we've built a science-grounded stack already outperforming the largest frontier models on the benchmarks that matter for trust. At a fraction of the compute cost 😎
Aloe retweeted
Yesterday's panel by @Vancity/@FrontierBC on "The AI Question: What do we actually want from it and who gets to decide?" -- moderated by Gurpreet Jhaj, with @GaryMarcus, @MadisonMills22 (@axios), and @arunbahl (@AloeInc) -- cut right to the heart of the question of how humans will exert or hand over our agency in this technological transition.
Closing remark couldn't have put a finer point on it:
Gurpreet: "What is one thing you want the founders and investors in this room to do differently when it comes to AI starting today?"
Arun: "You still have a lot of choice. The other machine that directs our behavior is the economy. Yet humans can exercise their decision making, and build things in line with their own values. So do that."
Aloe retweeted
Selling ads as a biz model is fundamentally misaligned with user interests — it undermines trust. How ya gonna trust AI that’s motivated to sell you stuff?
At @AloeInc user trust is the primary goal ➡️ some biz models are off the table.
Proudly. For the good of our species.
Aloe retweeted
It was a pleasure speaking with @PuckNews' A.I. correspondent @IKrietzberg about how @AloeInc's AI leapt ahead to state-of-the-art – and in doing so ushered in the “Dawn of the Self-Building A.I.”
We’ve taken a fundamentally different approach to AI because we are a fundamentally different kind of company: designed from the ground up to support human minds, not exploit them for their attention. We believe this is a prerequisite for AI we can trust.
Aloe retweeted
Spot on. @AloeInc proved exactly this a few weeks ago - a neurosymbolic system, designed and coded by @arunbahl, a cognitive scientist turned AI researcher (like @GaryMarcus!) - made a dramatic leap in capability over OpenAI et al on the GAIA benchmark for generalist agents.
Aloe retweeted
Critical convo to be having right now @StevenBartlett @DocAmen @sejnowski 🧠We are building @AloeInc from similar insights - knowing that a human’s clarity of thought will be our defining characteristic in the near future. AI that helps us think - rather than replaces our thinking - is critical for our species’ future.
You can watch the full conversation here: youtu.be/5wXlmlIXJOI
Aloe retweeted
“Evolutionary rather than revolutionary”… 😅
Except the revolution was happening where no one was looking - @AloeInc created an AI that literally builds itself.
First shot - easily leapfrogged the (now formerly) dominant players on the GAIA benchmark (ie tasks similar to day to day knowledge worker)
@GaryMarcus @slavov_n @DavidSacks
Kudos to the cognitive scientist/AI researcher @arunbahl for never being swayed by the hype and just building constantly from first principles.
This feels like a turning point for the AI hype.
Increasing number of influential AI enthusiasts, like @DavidSacks, are acknowledging the unrealistic narrative of super human intelligence and converging to the expectations articulated by @GaryMarcus.
Aloe builds itself. We are now state-of-the-art on the GAIA benchmark of generalist AIs, beating OpenAI, Manus, and Genspark by a wide margin.
How?
Like other AIs, Aloe uses tools. Unlike other AIs, if Aloe doesn't have the right tool for the job, it creates a new one first. This is composable program synthesis, and it's a new species of AI (we nicknamed it 𝘈𝘭𝘰𝘦 𝘩𝘢𝘣𝘪𝘭𝘪𝘴): when one Aloe increases its capability, they all get more capable together as they share tools and use them to create even better tools.
Aloe’s lead over other systems is highest on the most difficult scenarios - there’s plenty of headroom to expand.
We are just getting started – this is the floor of what we can do.
Aloe retweeted
Absolutely nailed it @om:
"However, in this new AI-first internet era, AI is your attention manager. So how does Meta translate its past business model of “capture and monetize attention” to “optimize and enhance attention?” The attention economy business model of “endless scroll for ad revenue” fundamentally breaks in the new AI reality."
This is why @arunbahl and I started @AloeInc - we fundamentally want to help people reclaim their agency - by equipping them with superhuman attention - and we are certain that the companies that built the Distraction Economy have 0% credibility toward that goal.
This is the post you want to read about Zuck’s Superintelligence Memo.
@om breaks it down and teaches us a thing or two about corporate communications.
