( the signal in the noise )

Technology moves fast. We move faster.

flux/ tracks the latest improvements across AI, silicon, energy and space — distilled into sharp, fast reads for people who build the future.

Read the pulse → Why speed matters

( 00 — on the bench )

flux/ FX‑1
flux/ FX‑1 — reference board 2nm — 48GB — 320W

scroll — it turns · click — enter the bench

( 01 — latest )What just got better

Four fronts where the state of the art moved this week.

/01Intelligence

Models that reason before they answer

Frontier models now plan, verify and self-correct mid-thought — trading raw speed for answers that hold up under pressure.

Read →
/02Silicon

2nm chips and the new power curve

Gate-all-around transistors are shipping, bending the performance-per-watt curve upward for the first time in years.

Read →
/03Energy

Grid-scale batteries get boring — finally

Storage costs keep sliding while capacity compounds. Boring is what winning looks like in infrastructure.

Read →
/04Space

Reusable rockets, weekly cadence

Launch is becoming a schedule, not an event. Orbit is turning into a place you ship to.

Read →
Translucent neural lattice with soft green nodes
Macro silicon wafer with a green circuit trace
White grid-scale battery cabinets in soft fog
Matte white rocket against a near-black sky
( 02 — the fronts ) 01 / 04
Machines learn to reason.
Intelligence — models that plan, verify, self-correct.
Silicon bends the curve.
2nm chips — more performance for every watt.
Energy gets boring.
Grid-scale storage — compounding quietly in the fog.
Orbit on a schedule.
Space — launch becomes logistics, not spectacle.
( 03 — why speed matters )
The half-life of tech news is 48 hours. If you read it slowly, you read it too late.
0×

more compute per watt across a single hardware decade.

0%

of new production code now written with AI assistance.

0+

orbital launches per year — and climbing every quarter.

( 04 — the signal ) field: noise

Chaos becomes signal.

/ 01 — static

The raw feed

Thousands of headlines a day. Unfiltered, tech news is just scattered noise.

/ 02 — the pulse

Fast tech news

flux/ compresses the feed into one sharp strike — speed is the product.

/ 03 — the curve

Everything compounds

Chips, code, capacity — the improvements bend up and to the right.

/ 04 — the orbit

Signal, on schedule

Launch cadence, release trains, weekly briefs — chaos settles into orbit.

/ 05 — the slash

All of it, in one stroke

flux/ — the whole feed, compressed into a single mark. Keep clicking to loop back.

Stay ahead of the curve.

One sharp brief on the latest improvements in tech. No noise, no filler — just the pulse.

Subscribe to flux/ →

( 01 — latest )

Sharp, daily briefs on what just got better in tech — the feed is warming up.

( the daily stack )
/01

Context windows cross 100M tokens

Long-context models stop summarizing your data and start remembering it.

intelligence — today /02

Solid-state batteries hit the road

First production EVs ship with solid-state packs — double density, half the charge time.

energy — yesterday /03

Three orbital launches in one week

Reusable boosters turn launch from an event into a timetable.

space — this week

( 02 — the fronts )

Intelligence, silicon, energy, space — the four arenas where the state of the art moves every week.

/01 — intelligence
/02 — silicon
/03 — energy
/04 — space
/01 — intelligence

reasoning models, agents, open weights — machines that plan before they answer.

/02 — silicon

2nm nodes, advanced packaging, memory — the substrate everything runs on.

/03 — energy

grid-scale storage, solar economics, fusion milestones — the power behind it all.

/04 — space

launch cadence, constellations, in-orbit compute — infrastructure above the sky.

( 03 — why speed matters )

Why a brief that arrives fast beats an essay that arrives late.

The half-life of tech news is 48 hours. If you read it slowly, you read it too late.

12×

more compute per watt across a single hardware decade.

54%

of new production code now written with AI assistance.

180+

orbital launches per year — and climbing every quarter.

01

Brevity is respect.

Every brief earns its length. If it can be a sentence, it is a sentence.

02

Signal over noise.

We skip the hype cycle and track what actually shipped, benchmarked, and scaled.

03

Fast, with proof.

Speed never means sloppy — every claim points back to a primary source.

( 04 — the signal )

A living field of 420 dots that folds chaos into five meaningful shapes.

noise → pulse → curve → orbit → flux

Play with the field →
/01noise

thousands of scattered headlines a day — the raw feed.

/02pulse

the beat of what actually matters, found in the chaos.

/03curve

compounding progress, bent upward and to the right.

/04orbit

four fronts in steady rotation around one reader: you.

/05flux/

the whole field, distilled down to a single slash.

( 05 — subscribe )

One sharp brief a week. No noise, no filler — just the pulse.

( one brief, straight to you )
/ one brief, every week / a five-minute read / zero spam, zero filler
x / twitter github linkedin instagram rss contact us

( unsubscribe anytime — your inbox stays clean )

( 06 — the bench )

Two exhibits of state-of-the-art silicon — a bare GPU, and the machine a chip lives in. Pick one — scroll, and it descends with you while the spec sheet draws itself on.

GEFORCERTX 5090
NVIDIA5090
GeForce RTX 5090/ gpu

Blackwell at full tilt — 92 billion transistors and light that behaves.

step onto the bench →
apple siliconM5APL1M05 — 3NM
MacBook Pro M5 Max/ laptop

Twenty-eight billion transistors, dressed in aluminum — the machine built around the chip.

step onto the bench →

( bench /01 — gpu )

← back to the bench  ·  ( 3d model — mistjs / sketchfab, cc by-nc )

5090
GEFORCERTX 5090
NVIDIA5090
/01 — architecture

92 billion transistors on TSMC 4NP — the largest consumer die ever shipped.

/02 — memory

1.79 terabytes per second of bandwidth — frames come from memory, not miracles.

/03 — compute

Fourth-generation ray tracing, fifth-generation tensor — light finally behaves.

/04 — power

Double flow-through fans over a two-slot vapor chamber — heat is a design problem.

( end of exhibit — next: macbook pro → )

( drag to spin ) ( 000 / 100 )

( bench /02 — the machine )

← back to the bench  ·  ( 3d model — techfreak / sketchfab, cc by )

MAX
apple siliconM5APL1M05 — 3NM
/01 — process

Third-generation 3nm silicon — twenty-eight billion transistors at the quiet heart of the machine.

/02 — memory

One pool at 153 GB/s — CPU, GPU and neural engine share the same thought.

/03 — neural

A 16-core neural engine that treats AI as a native tongue, not a peripheral.

/04 — efficiency

Desktop-class performance at whisper power — the fan is a formality.

( end of exhibit — next: rtx 5090 → )

( drag to spin ) ( 000 / 100 )