Tyler Benner

I beat the track up like a hundred arms
I'm funky like underarms

[Tyler Benner @tyler_benner]

This second ad highlights speed.

Spreading your toes, feeling the pedals, and staying grounded through corners improves the riding experience. It boosts confidence, enhances balance, and allows for instant acceleration.

More is coming, but enjoy the ride now.

🚴🚴‍♂️💨💨

[Fred Chen @DrFrederickChen]

Here we examine the stochastic effects of EUV lithography for 32 nm line pitch, where, unlike metal interconnects, the lines are narrower than the half-pitch, as in DRAM word lines and bit lines. There is a significant impact on edge roughness. https://t.co/e5upru8byM

[포시포시 @harukaze5719]

Vertical Stacking of Transistors Themselves

Previously, when thinking of 3D stacking, one would have imagined stacking manufactured chips.

However, the result Samsung demonstrated in its VLSI this time involves 3D stacking the N- and P-type transistors themselves, which form the foundation of semiconductors.

By 3D stacking N- and P-type transistors while maintaining the three sheets of the existing GAA, it is theoretically possible to achieve double the density and higher power efficiency.

In addition, a Gate Pitch of 42nm was achieved alongside this.

[Satya Nadella @satyanadella]

https://t.co/vLmiBKTtX3

[John Scott-Railton @jsrailton]

NEW: malware developers added nuclear & biological weapons text to to their spyware.

Goal? To trigger LLM safety refusals... so that their spyware wouldn't be analyzed by an AI security scanner.

Cleanest practical example I can think of for why over-indexing on first order safety alignment is risky.

When closed (and open) models ship with aggressive refusals, they will be sprinkled with second-order blindspots that attackers will discover...and exploit.

We are only in the earliest days of attackers leveraging these features, and it wouldn't surprise me if users systems that need to handle complex cybersecurity issues demand that models be less safety-blunted.

In the weeds: @SocketSecurity's post also shows why intention matters in how you design a malware analysis pipeline to avoid prompt manipulation.

H/T to colleagues that shared this with me https://t.co/f3Aj9TYxU4

[Fred Chen @DrFrederickChen]

Stochastic effects are the ultimate limiters of EUV lithography. Even OPC is subject to stochastic failure. KLA found that 38 nm pitch metal in an SRAM cell shows bridging failure at 5 sigma for a metal oxide resist. https://t.co/eUFWRjv62L

[🌿 lithos @lithos_graphein]

Intel 18A is a game changer. This is an absolutely amazing TEM image of the nanosheet & power via patterning at play here in this device. Flawless.

[The Dr. Logos © @TheDoctorLogos]

The meaning behind The Shadow (1909)

[Hassan Mafi ‏ @thatdayin1992]

Americans realizing they can go to jail for smoking weed but Trump gets a $200 fine for insider trading

[CITIUS MAG @CitiusMag]

“Today, I took two slices of bread, honey and tea.” 🥖 🍯 🫖

Sabastian Sawe revealed what he had for breakfast before breaking the marathon world record in 1:59:30 at the London Marathon.

🎥 Full post-race press conference available to watch on The CITIUS MAG YouTube channel.

[Black Hole @konstructivizm]

The reason why large asteroids don't fall to Earth every day and cause disasters is because Jupiter's gravity attracts asteroids and protects the inner planets.

[Bearly AI @bearlyai]

Nvidia’s Chief Scienst Bill Dally tells Jeff Dean how Nvidia uses AI to speed up the chip design process:

▫️trained an LLM on all proprietary internal Nvidia docs over past 30+ years (junior employees query it instead of interrupting senior designers)

▫️one AI tool ports Nvidia’s cell library to a new semiconductor process and does it in only one night (used to take 10 employees up to 8 months, or 80 total person-months)

▫️since the 1950s, there’s been a classic chip design problem of where to place look ahead stages in a chain (AI is coming up with solutions using “bizarre designs thay no human” would think of)

▫️agentic AI systems are doing a ton of exploration…testing parameters spaces…suggesting new approaches…running architecture experiments

▫️verification process is laborious but AI able to do it at fraction of time (compresses time from design to tape-out, with TSMC making chip)

Dally says it’s still a long way from end-to-end chip design but imagines a world where one master AI agents manages multiple sub-agents (similar to the current human-led process).

[delian @zebulgar]

Ok this is the best one holy shit

[Epic Maps 🗺️ @theepicmap]

This is why we always see the same side of the Moon and its phases

[NASA @NASA]

Hello, Moon. It’s great to be back.

Here’s a taste of what the Artemis II astronauts photographed during their flight around the Moon. Check out more photos from the mission: https://t.co/rzM1P0QbOl

[NASA @NASA]

We see our home planet as a whole, lit up in spectacular blues and browns. A green aurora even lights up the atmosphere. That's us, together, watching as our astronauts make their journey to the Moon.

[Nav Toor @heynavtoor]

🚨BREAKING: OpenAI published a paper proving that ChatGPT will always make things up.

Not sometimes. Not until the next update. Always. They proved it with math.

Even with perfect training data and unlimited computing power, AI models will still confidently tell you things that are completely false. This isn't a bug they're working on. It's baked into how these systems work at a fundamental level.

And their own numbers are brutal. OpenAI's o1 reasoning model hallucinates 16% of the time. Their newer o3 model? 33%. Their newest o4-mini? 48%. Nearly half of what their most recent model tells you could be fabricated. The "smarter" models are actually getting worse at telling the truth.

Here's why it can't be fixed. Language models work by predicting the next word based on probability. When they hit something uncertain, they don't pause. They don't flag it. They guess. And they guess with complete confidence, because that's exactly what they were trained to do.

The researchers looked at the 10 biggest AI benchmarks used to measure how good these models are. 9 out of 10 give the same score for saying "I don't know" as for giving a completely wrong answer: zero points. The entire testing system literally punishes honesty and rewards guessing.

So the AI learned the optimal strategy: always guess. Never admit uncertainty. Sound confident even when you're making it up.

OpenAI's proposed fix? Have ChatGPT say "I don't know" when it's unsure. Their own math shows this would mean roughly 30% of your questions get no answer. Imagine asking ChatGPT something three times out of ten and getting "I'm not confident enough to respond." Users would leave overnight. So the fix exists, but it would kill the product.

This isn't just OpenAI's problem. DeepMind and Tsinghua University independently reached the same conclusion. Three of the world's top AI labs, working separately, all agree: this is permanent.

Every time ChatGPT gives you an answer, ask yourself: is this real, or is it just a confident guess?

[TSN @TSN_Mad]

@LookAtMyMeat1 Enough for him to betray everything he campaigned on. Trump is a traitor to the American people and will go down as the worst president because of that.

[🌿 lithos @lithos_graphein]

While everyone's distracted by the latest drama online, let's talk real progress in chip lithography this week:

Inkjet Adaptive Planarization (IAP).

This process delivers ultra-flat wafer surfaces essential for focus-sensitive patterning—especially with High-NA EUV, where depth of focus shrinks dramatically.

Canon adapts its nanoimprint lithography tool in reverse: inkjet dispenses a light-curable material (the "ink") precisely matched to the wafer's existing topography, then presses a superflat quartz plate to level everything in one go.

Result? Single-digit nm flatness (~5 nm uniformity) across a full 300 mm wafer, regardless of underlying circuit density or pattern variations. That's game-changing precision for downstream exposures.

Key players:

- Canon: the IAP tool

- Shin-Etsu: the superflat quartz substrate

- EMD: the planarization ink

[el.cine @EHuanglu]

AI imagine Hawking vs Newton.. seedance 2 got big brain

[Benjamin Bardou @benjaminbardou]

Revolution
[a complete circle or turn round a centralpoint, axis]
[to roll back, to bring back]