OpenAI’s public GitHub engineering output accelerated in August: 3.96M lines changed and 3,543 commits vs. Anthropic’s 1.32M and 2,785. Public GitHub repository activity provides a directional "proxy" for engineering productivity.

Korea Semiconductor Exports | June 2026 [Export Value] - DRAM (including modules): Export value of $21,846mn, Y/Y +385%, M/M +17% - DRAM (excluding modules): Export value of $11,176mn, Y/Y +388%, M/M -2% - Flash memory (NAND): Export value of $2,486mn, Y/Y +301%, M/M +44% - MCP (HBM): Export value of $12,681mn, Y/Y +171%, M/M +32% - SSD: Export value of $5,158mn, Y/Y +355%, M/M +30% [Export Unit Price] - DRAM (excluding modules) export unit price was $74,687/kg, up +514% Y/Y and down -4% M/M - DRAM (including modules) export unit price was $59,717/kg, up +501% Y/Y and down -2% M/M - Flash memory export unit price was $62,115/kg, up +451% Y/Y and +9% M/M - MCP (HBM) export unit price was $94,132/kg, up +115% Y/Y and +12% M/M - SSD export unit price was $21,131/kg, up +309% Y/Y and down -5% M/M

Sadly I couldn’t make it to the Cerebras event today b/c I was having fun with my friends. But this is what I gather… The new WSE-3T is on the same N5 node and same wafer size but doubled compute to 250 PF & doubled memory bandwidth to 43.2 PB/s which to me would mean a 2x clock increase which is minimum 2x the power. The prior WSE-3 consumed ~23 kW so let’s call it 40-55 kW per wafer. 3 wafers per rack or 130-180 kW for the entire system. Rubin NVL72 is 72 packages x 288 GB of HBM4 @ 22 TB/s for a total of 1.6 PB/s at roughly ~200 kW. So for CS-4 system we get ~0.9 PB/s per kW vs. ~8 TB/s per kW on Nvidia. So roughly 100x more bandwidth per kilowatt and in practice it’s probably more cause Nvidia relies more on interconnect than Cerebras. Sooooo prettay prettay good 👍 But only 132 GB of SRAM vs. ~21 TB of HBM memory + 75 TB of LPPDR on Nvidia it’s a lil scaaaary. I’m all for a decode optimized accelerator but when you can’t fit a Qwen 3.8 or DeepSeek V4 on it I don’t know how much of a value add it really is. If I still need to buy an Nvidia rack to go along with my Cerebras rack why would I really bother? Fast tokens spend faaaaast money 💰!!! Good if you’re rich but sadly most of us are working class token consumers who can’t afford a Bugatti model. That being said, the I/O board is a great addition (does it have FPGA???) and I don’t think the 7.2 TB/s is a big deal as latency is more important since this will only be used for inference. Also modular power solutions hmm .. 🤔 nice touch ! 😏 I see the vision. Giving Cerebras a B+ and I am excited to see what they do next, I can tell they are listening to some of the advice we gave them! Now let’s see if they are willing to diversify a bit so they can get those gross margins up. There’s still time to turn the boat 🛶 around!

JUST IN Zurich discounts Tesla insurance for FSD (Supervised) — Australian first. • InsureMyTesla now rates FSD as lower risk • Tesla data: 7× fewer major/minor collisions vs regular EVs • 132M km FSD already driven in Australia • Second market worldwide to price FSD this way

ONLYFANS BY THE NUMBERS • 5,076 creators have earned $1M+ • $30B+ paid to creators since 2016 • $6.3B paid out in FY25 alone • 5M creator accounts, 2.5M active • 437M registered fans, 132M active • OnlyFans keeps a 20% cut • Just 47 employees in FY25 • Valued recently at $3.15B

Gracie Abrams debuts 6 songs off her new album ‘Daughter from Hell’ on the Global Spotify chart: #49#. Good Reason — 2.05M #75#. Look at My Life — 1.69M #95#. Hit the Wall — 1.57M #143#. Death Wish — 1.36M #160#. The Knife — 1.32M #196#. Daughter from Hell — 1.23M

132 days until the GENIUS Act takes effect. Regulators missed their own July deadline for final rules. So issuers are trying to build compliance systems without a rulebook. One thing holds no matter how the rules land. You have to prove where your onchain data came from.