Megha Jain

12 followers · 7 following · 2030 views

on the atlas — 38

highlights — 290

this method still makes too many word-level or character-level errors to be practical. Still, our results follow a scaling law: the more data used for training, the better the decoder, without -- at least for now -- a detectable performance plateau.
Brain2Qwerty — Decoding typed sentences from non-invasive brain activity
reaches up to 78% word accuracy for the best participant.
Brain2Qwerty — Decoding typed sentences from non-invasive brain activity
non-invasive magnetoencephalography (MEG) recordings
Brain2Qwerty — Decoding typed sentences from non-invasive brain activity
the first such designation ever applied to an American company
Anthropic Supply Chain Risk Designation Takes Effect — Latest Developments and Next Steps for Government Contractors | Insights | Mayer Brown
a structured, replayable history of how context turned into action
AI’s trillion-dollar opportunity: Context graphs - Foundation Capital
If you persist those traces, you get something that doesn’t exist in most enterprises today: a queryable record of how decisions were made.
AI’s trillion-dollar opportunity: Context graphs - Foundation Capital
Agents don’t just need rules. They need access to the decision traces that show how rules were applied in the past, where exceptions were granted, how conflicts were resolved, who approved what, and which precedents actually govern reality.
AI’s trillion-dollar opportunity: Context graphs - Foundation Capital
Ball’s framing assumes the data agents need already lives somewhere, and agents just need better access to it plus better governance, semantic contracts, and explicit rules about which definition wins for which purpose.
AI’s trillion-dollar opportunity: Context graphs - Foundation Capital
anything that touches the outside world is an activity, and anything that only coordinates and makes decisions is the workflow.
Temporal Primer - Building Long-Running Systems
Spawn goroutines or threads that run outside Temporal’s control
Temporal Primer - Building Long-Running Systems
Make direct network calls or database queries - those belong in activities
Temporal Primer - Building Long-Running Systems
The replay mechanism requires that given the same history, the workflow always makes the same decisions.
Temporal Primer - Building Long-Running Systems
Temporal handles retrying Activities automatically when they fail.
Temporal Primer - Building Long-Running Systems
Activities are the “do a thing” units. They are explicitly not deterministic
Temporal Primer - Building Long-Running Systems
Workflow functions must be deterministic
Temporal Primer - Building Long-Running Systems
The key aspects to his approach seem to be to articulate what you want to have happen, develop any sort of approach to starting to tackle it, and then act on that approach.
Being Someone who Does Things - by Joe Hovde
This becomes more interesting over the course of a decade or so, because you can clearly see the compounding benefits of actively productive activities, and you can see too much leisure have real negative effects
Being Someone who Does Things - by Joe Hovde
they are expected to generate additional property taxes equal to one third of Abilene’s current total property tax base
Inside Stargate
the entire cooling system across all eight buildings only needs to be topped off each year with an amount of water equal to that used by four households per year
Inside Stargate
Our available compute grew roughly 3x year-over-year from 0.2GW in 2023, to 0.6GW in 2024, to ~1.9GW in 2025
Inside Stargate
i don’t think hope is something to be embarrassed about. i think it’s actually one of the bravest things about us
you don't get to hate the girl that got you here
When I first read it, I took it to mean that we are all composites of the world around us, that the self is assembled from external material. Now I believe the inverse is also true. The world is shaped by an internal vitality, by what I do when no one is watching.
making yourself proud - starting from nix
That is the very underpinning of transformation: you can’t predict how it will warp you or turn you into who you weren’t before. It simply feels like acute recognition at the site of change: the way you recognize a perfect fit on a pair of jeans, or begin to fall for someone. It’s just like, Oh.
choice loneliness - starting from nix
So here’s my advice: stop trusting == when wrappers are involved
Interviewer asked me why is 1 == 1 True but 1000 == 1000 False in Java? | by Harry | JavaGuides | Medium
Use .equals() like a sane developer.
Interviewer asked me why is 1 == 1 True but 1000 == 1000 False in Java? | by Harry | JavaGuides | Medium
Java maintains a cache of Integer objects from -128 to 127. That’s it.
Interviewer asked me why is 1 == 1 True but 1000 == 1000 False in Java? | by Harry | JavaGuides | Medium
It hands you the same cached Integer instance. Both x and y are literally pointing to the same memory reference.
Interviewer asked me why is 1 == 1 True but 1000 == 1000 False in Java? | by Harry | JavaGuides | Medium
Imagine yourself as the person you would be afraid to say you want to be out loud to others
Becoming a magician – Autotranslucence
Sometimes, jumping ship and having no strategy for a while can be better, and allow you to clarify what you want
Becoming a magician – Autotranslucence
often in areas when you are actually making progress you won’t be able to imagine a nonlinearly better version of yourself, only one who in fact followed the current strategy to its logical conclusion and is now about as great (at the thing the strategy is for) as you expected to be
Becoming a magician – Autotranslucence
asking my most interesting friends to introduce me to their most interesting friends, going down similar rabbit holes with the bibliographies of books that excite me, and generally living in ‘explore’ mode at various points in life
Becoming a magician – Autotranslucence
Magicians or their work often seem to have a subconscious glow that I am drawn to, particularly if they use a type of magic that I recognise is on my critical path and thus something I’m currently seeking
Becoming a magician – Autotranslucence
The ‘describe the version of you that seems impossible right now’ trick I described above is largely an attempt to bypass that part of my brain that dismisses the work of magicians as crazy and starts allowing it to make the necessary shifts required to become the kind of magician I am envisioning.
Becoming a magician – Autotranslucence
And that, in fact, is my definition of magic – competence so much more advanced than yours with such alien mental models that you cannot predict the outcomes of the model at all.
Becoming a magician – Autotranslucence
improvement on tasks of different complexity has varied: models are now cheaper to use for the most complex tasks, but only somewhat cheaper for simple tasks
Deep Dive: Economics of the AI Build-Out | Contrary Research
there is an imbalance between the present demands of data centers and the compute capacity they can provide
Deep Dive: Economics of the AI Build-Out | Contrary Research
Both upgrades to existing data centers and construction of new data centers are major expenses for the developers of AI models, either directly or via compute partners
Deep Dive: Economics of the AI Build-Out | Contrary Research
The part of the equation this narrative fails to capture is the increasing demand for AI capabilities
Deep Dive: Economics of the AI Build-Out | Contrary Research
remains largely optimistic given the falling cost of inference, with OpenAI’s March 2025 letter to Congress noting that the cost of a given AI capability falls by an order of magnitude every 12 months
Deep Dive: Economics of the AI Build-Out | Contrary Research
no frontier AI company relying on third-party data centers is yet profitable
Deep Dive: Economics of the AI Build-Out | Contrary Research
rationalize AI demand
The AI Boom Needs a Market for Compute, Just Like Oil and Spectrum - Bloomberg
That in turn maximizes value, not least because it significantly reduces the risk associated with buying compute.
The AI Boom Needs a Market for Compute, Just Like Oil and Spectrum - Bloomberg
Compute markets — both spot and futures — involve so many variables that they’ve become a possibility only with the advent of AI agents that can accept natural-language inputs
The AI Boom Needs a Market for Compute, Just Like Oil and Spectrum - Bloomberg
when chunks of the electromagnetic spectrum previously allocated by governments started being auctioned, priced and invested in
The AI Boom Needs a Market for Compute, Just Like Oil and Spectrum - Bloomberg
while vendors enjoy the extra pricing power the status quo gives them, they’d also love to be able to compete effortlessly on every contract being negotiated anywhere in the world
The AI Boom Needs a Market for Compute, Just Like Oil and Spectrum - Bloomberg
Walmart Inc. after all, is a lot more profitable than LVMH Moët Hennessy Louis Vuitton SE
The AI Boom Needs a Market for Compute, Just Like Oil and Spectrum - Bloomberg
compute providers will be more akin to Walmart, with its high-volume, low-margin stores, than a high-margin, medium-volume business like LVMH
The AI Boom Needs a Market for Compute, Just Like Oil and Spectrum - Bloomberg
What’s unclear is whether those valuations are based on a scarcity thesis — that demand is likely to outstrip supply for years to come — or an abundance thesis, foreseeing a multitrillion-dollar market where these companies will be among the many winners.
The AI Boom Needs a Market for Compute, Just Like Oil and Spectrum - Bloomberg
as long as they agree to be curtailable or colocate with dispatchable generation.
Unpacking DOE's proposal to transform data center interconnection | Latitude Media
To fund this, tech giants are turning to a growing list of complex debt financing options, including corporate debt, securitization markets, private financing and off-balance sheet vehicles
Is OpenAI Already Too Big to Fail?