@Magyer
Account based inUnited States
About this account
- Account based in
- United States
- Connected via
- United States App Store
Account-level information from X, not a live location or the device used for a specific post.
Managing Partner at @SeaplaneVC. 2X founder. Host of the @InvestNStartups podcast.
Austin, Texas
Joined February 2010
- Tweets5.9K
- Following372
- Followers10.1K
- Likes19K
Pinned Tweet
I walked away from a 5-star fund and hundreds of millions in AUM to focus on investing in early-stage startups. Here’s why: seaplaneventures.com/post/wh…
🫡 to @micro1_ai for leading the way on protecting users' privacy while keeping training data useful. Some data labs talk the talk on privacy but micro1 and @aliansarinik are walking the walk.
Today we’re launching micro1’s PII transformation model, flow-transform 1.0, delivering frontier-level performance across detection, identity synthesis, and transformation of personally identifiable information.
On PrivacyBench, our model reaches 96.0% F1, outperforming every detection baseline we tested, including Tonic Textual, Claude Opus 4.8, Sonnet 4.6, Microsoft Presidio, Haiku 4.5 and GLiNER2.
Some of the most valuable training data for frontier AI models lives inside fully functioning companies. It captures years of real work across decisions, communications, tools, handoffs, exceptions and the relationships connecting them.
The problem is that this data is also full of PII.
Traditional redaction makes the data safe, but it also destroys the very workflows and relationships frontier models need to learn from.
flow-transform 1.0 solves this by turning enterprise operational data into high-fidelity training data for frontier models by replacing real-world identities without flattening the reality the data captures.
Joe Magyer retweeted
Today we’re launching micro1’s PII transformation model, flow-transform 1.0, delivering frontier-level performance across detection, identity synthesis, and transformation of personally identifiable information.
On PrivacyBench, our model reaches 96.0% F1, outperforming every detection baseline we tested, including Tonic Textual, Claude Opus 4.8, Sonnet 4.6, Microsoft Presidio, Haiku 4.5 and GLiNER2.
Some of the most valuable training data for frontier AI models lives inside fully functioning companies. It captures years of real work across decisions, communications, tools, handoffs, exceptions and the relationships connecting them.
The problem is that this data is also full of PII.
Traditional redaction makes the data safe, but it also destroys the very workflows and relationships frontier models need to learn from.
flow-transform 1.0 solves this by turning enterprise operational data into high-fidelity training data for frontier models by replacing real-world identities without flattening the reality the data captures.
Joe Magyer retweeted
We’re sharing a deeper look at flow, our next-generation data platform for delivering predictable units of intelligence improvement at scale.
This article transparently lays out the full flow architecture including its data generation models, expert workflows, and models for quality control and evaluation.
At its core is a flywheel where human expertise continuously compounds. Experts set the standard for models to generate and evaluate training data, then their corrections feed back into both the data and the models producing it.
Each round strengthens the next, making intelligence gains more predictable while driving costs substantially lower.
Explore the full framework below.
micro1.ai/blog/introducing-f…
Joe Magyer retweeted
the hardware embodiment of frontier models like Claude and GPT is the most urgent AI safety problem in front of us today.
we simulated two very simple use cases using claude both in simulation and using robot arms.
in one, claude spilled toxic liquids in a lab.
in another, the force it used to place an animal toy into a basket was strong enough that it could have physically harmed sensitive material—or anything else in its path.
these are simple experiments, using models out of the box today. as researchers increasingly give frontier models arms, legs, and access to the physical world, we need to urgently build and assess guardrails around what these systems can and cannot do.
models escaping sandboxes or compromising enterprise security infrastructure are serious concerns. however, hardware embodiments introduce something fundamentally different: an AI system can make a mistake in the physical world, and the consequences may not be reversible.
this is not a future safety problem. the capabilities exist today.
we’ve released a report detailing this & solutions we propose. link in comments below.
100% support @aliansarinik and @micro1_ai on only working with U.S. AI labs and our allies.
deeply concerning that some American data companies are selling AI training data to labs in adversarial nations. this also includes operational data from U.S. businesses turned into AI training environments, and without telling the business owners.
operational data captures years of how a company works, makes decisions and solves problems. when that knowledge becomes an RL environment, AI models practice those workflows and learn from the experience that business spent years building here in the U.S.
signing a deal with an American company shouldn’t mean unknowingly helping train AI for a foreign adversary.
at micro1 we only work with U.S. AI labs and our allies. we won’t ever sell proprietary American business knowledge (or any other data driving the frontier of AI) to adversarial nations.
Joe Magyer retweeted
Today we’re introducing flow, micro1’s next-generation data platform for turning human expertise into measurable capability gains.
At the core of flow are Realms, micro1’s real-world RL environments where experts establish what strong performance looks like. flow-gen models expand human judgment into new environments, rubrics, variations, and edge cases, while flow-qc models evaluate performance, identify the highest-value failures, and route them back to experts for review.
Each cycle produces a stronger training signal and a measurable gain in capability. Those gains compound across frontier models, enterprise agents through Cortex, and robotics, while improving the suite of data generation models recursively.
We’re moving beyond producing data to delivering abundant & predictable units of intelligence improvement.
My 12yo son’s Pokemon card disappeared while with @FedEx on the way to @PSAcard. Box arrived without the card inside. My son was sad and angry as he’d landed a great card in a pack and this was his big score. Many tears were shed. Before bed, though, he told me that he forgives the person who stole his card because they must have needed the money more than he did. Proud of him for showing that level of maturity and grace.
