WΞNDΞL

Microsoft is preparing a major Windows 11 performance push, and the real target is WinUI. Windows 11 is moving to WinUI, including the Start menu and Agenda view in Notification Center. At Build 2026, Windows UI engineer Chris Anderson previously said Microsoft has invested heavily in reducing WinUI memory usage and moving it to the Windows system compositor for better performance. Right now, you can feel the WinUI tax in places like File Explorer. Open “This PC” and then compare it with the WinUI-based Home tab. The difference is obvious, especially on low-end hardware. Microsoft is testing memory, performance, and reliability fixes before rolling out the WinUI Start menu. The UI may not look dramatically different, but it should stutter less and use less RAM. Microsoft is fully committed to major Windows 11 performance improvements.

I wrote about why we believe the future is for everyone. More coming about a positive vision for a world with superintelligence soon.

GrapheneOS is cool and badass

Moderate alcohol consumption is... good for you? The research will shock you. Let's take a look ⬇️

JUST IN: Pentagon moves to build hyperscale AI data centers on at least a dozen U.S. military bases.

Type 0.1 + 0.2 into Python, and it spits out: 0.30000000000000004. COBOL was built to never make that error, but its flawless financial math makes it surprisingly clumsy with fractions. COBOL was created specifically for business data processing. Its numeric model centers on fixed-point decimal arithmetic rather than the binary floating-point arithmetic used by most other languages. Money is almost always expressed in decimal (dollars and cents, or the equivalent in other currencies). COBOL therefore stores and calculates numbers as true decimal digits. The usual storage form is packed decimal (COMP-3 / Binary Coded Decimal). Each digit occupies four bits, and the scale (the position of the decimal point) is fixed by the programmer in the PICTURE clause—for example, PIC S9(7)V99 means seven digits to the left of the decimal point and exactly two to the right. Arithmetic on these values is performed digit-by-digit in decimal. Consequently: •0.10 + 0.20 equals exactly 0.30 •Interest of $14.73 is stored and reported as exactly $14.73 •Rounding is under explicit programmer control with the ROUNDED phrase •Results are bit-for-bit reproducible across COBOL compilers These properties matter enormously in finance. Tiny binary floating-point errors (the classic 0.1 + 0.2 = 0.30000000000000004) compound across millions of accounts and can produce regulatory violations or reconciliation nightmares that take weeks to unwind. COBOL’s exact decimal math eliminates that class of error by design. The same fixed-scale decimal model is what makes COBOL comparatively weak with unrestricted fractions and real numbers: •- The number of digits after the decimal point must be declared in advance. There is no automatic “floating” scale. •- Representing a true fraction such as 1/3 or a high-precision intermediate result requires you to choose a large fixed precision (for example, PIC 9(5)V9(30)). That consumes memory and slows arithmetic. •- Very large or very small magnitudes, or numbers whose required precision varies widely during a calculation, force either truncation/overflow or the use of extremely wide fields. •- COBOL does support binary floating-point (COMP-1, COMP-2, and later IEEE formats), but these are secondary, less commonly used in business code, and still suffer the usual binary representation problems for decimal fractions. •- Scientific or engineering work that needs dynamic range, transcendental functions, or arbitrary intermediate precision is far more natural in languages whose default numeric type is binary floating-point (Fortran, C, Python, etc.). In short, COBOL’s strength and its limitation are two sides of the same coin. It deliberately chose exact, fixed-scale decimal arithmetic so that financial calculations would be precise, auditable, and free of binary rounding artifacts. That choice makes it the language of record for money, but it leaves the language less convenient for general fractional or scientific real-number work that expects a floating scale and binary floating-point hardware acceleration.