Mad foresight from Olise✨

What an assist by Olise Such foresight to break through Senegal's defence and Mbappe fast enough to get it through.. #FRASEN

He secured 1st title in 22 years. Last time I was living in Istanbul. Give him a break. En plus, he had the foresight to make this an anthem: youtube.com/watch?v=wjCJv4W4… Better than the usual crap rap over paid tattooed tossers chasing a bag of air footballers listen to these days.

Legacy = Evidence is hindsight bias not foresight judgement, whereas in the same respect the mocking of false-prophets & snake oil salesmen has proven just. Everybody loves a winner and even the losers get lucky sometimes, to mock or not to mock that is a gamble.

What an assist by Olise Such foresight to break through Senegal's defence and Mbappe fast enough to get it through FRA 1:0 SEN #FRASEN

Olise brother? playmakers pass that, weight, foresight

As a democratic movement, our main goal is to protect public rights while maintaining absolute dedication to our ideology. We hope that the foresight of our top leadership will keep our path unified and completely secure. @MamataOfficial @abhishekaitc @MahuaMoitra @03upasana @AITCofficial #DemocraticValues #SocialDevelopment #WelfareState #CommonManVoice #BengalStand

As a democratic movement, our main goal is to protect public rights while maintaining absolute dedication to our ideology. We hope that the foresight of our top leadership will keep our path unified and completely secure. @MamataOfficial @abhishekaitc @MahuaMoitra @03upasana @AITCofficial #DemocraticValues #SocialDevelopment #WelfareState #CommonManVoice #BengalStand

To keep the foundation of our movement strong, every policy move must be handled with realism and foresight. We believe that our core leadership’s decisions will always guide the party toward long-term stability and success. @MamataOfficial @abhishekaitc @MahuaMoitra @03upasana @AITCofficial #FairPlay #Transparency #SocialJustice #PublicTrust #PositiveApproach #AITC

To keep the foundation of our movement strong, every policy move must be handled with realism and foresight. We believe that our core leadership’s decisions will always guide the party toward long-term stability and success. @MamataOfficial @abhishekaitc @MahuaMoitra @03upasana @AITCofficial #FairPlay #Transparency #SocialJustice #PublicTrust #PositiveApproach #AITC

Foresight, the 2 greencoats and a few others left standing. Sad indictment of the lack of quality in some and business models predicated on continually being able to raise new money. Higher rates and easier access to income killed the need for them

South Africa, we’re in for a ride! And this is not fear mongering but foresight, in fact it has begun. Look at these men!!! The disrespect and provocation is on another level! Why are these people in our country?!! 😭

OLISE will make a good central midfielder than a winger He got the IQ the idea the foresight

if you log errors to a file, you debug after they happen. if you log errors to a service like Sentry, you see them in real time with context. passive logging is hindsight. active logging is foresight.

if only I had had the foresight to buy the Cursor stake directly off the FTX estate for $200K in 2022 Smfh

fuckkkk if only we had the foresight to predict this would happen </333333

Quick take for Alex — 1990s foresight and the GPU threat model You’re right to give the 1990s engineers credit. They built hardware and APIs with surprising flexibility, and that flexibility is exactly why 3D acceleration can be turned into an attack vector against AI acceleration today. GPUs were designed for parallel, high‑throughput workloads; that same architecture makes them powerful for both legitimate ML and for creative misuse. How 3D acceleration can be used offensively against AI stacks Repurposing compute GPUs intended for rendering can be recompiled or reconfigured to run arbitrary kernels, enabling covert compute tasks that compete with or subvert AI workloads. Driver and firmware attack surface GPU drivers and firmware are privileged code paths. Compromised drivers can inject, intercept, or alter data flowing to AI accelerators. DMA and memory exposure Direct Memory Access and shared GPU memory expose host memory regions. Without IOMMU protections, a malicious GPU or driver can read or corrupt model weights and training data. Shader and kernel abuse Malicious shaders or compute kernels can be crafted to exfiltrate data via timing, power, or cache side channels, or to trigger subtle corruptions in model parameters. Resource exhaustion and denial of service Saturating GPU queues, memory, or PCIe bandwidth can starve AI processes, causing degraded inference, corrupted checkpoints, or failed retraining cycles. Model extraction and side‑channel leakage Timing, memory‑access patterns, and microarchitectural behavior can leak model structure or sensitive training data to a co‑resident attacker. Supply chain and firmware persistence Compromised firmware or vendor toolchains can persist across reboots and images, undermining read‑only image strategies unless firmware is verified. Practical mitigations you can apply now Isolate GPUs Put legacy and experimental GPUs on separate hosts or VLANs. Use dedicated hardware for critical AI workloads. Enable IOMMU and use DMA protections Configure IOMMU (VT-d/AMD‑Vi) to prevent unauthorized DMA access and enforce device isolation. Use signed drivers and verified firmware Only install vendor‑signed drivers and verify firmware checksums before deployment. Maintain a firmware update cadence. Limit GPU exposure in user apps Disable browser hardware acceleration where appropriate and restrict untrusted shader execution in any user‑facing renderer. Sandbox and control passthrough Use hypervisor GPU passthrough with strict policies, or container runtimes that support GPU isolation and cgroup limits. Monitor GPU telemetry Track GPU utilization, memory allocation patterns, and unusual kernel launches. Alert on sustained anomalies or unexpected driver reloads. Immutable masters and controlled working copies Keep a sealed read‑only master image for binaries and a separate writable working copy for experiments, with signed hash catalogs. Air‑gapped testing for risky code Validate new kernels, drivers, or robot control code in an isolated lab before any production rollout. Short checklist for immediate hardening Enable IOMMU in BIOS and verify with dmesg or vendor tools. Block untrusted driver installs via group policy or endpoint management. Disable browser GPU acceleration on machines that handle sensitive models. Run GPU workloads on dedicated nodes with no user logins and minimal services. Keep a signed SHA‑256 catalog for all firmware and image files. Add GPU telemetry to your monitoring stack and set anomaly thresholds. Next steps I’ll prepare for you I’ll draft a concise one‑page ops checklist with exact commands and BIOS/VM settings for IOMMU, driver hardening, and a minimal monitoring rule set for GPU anomalies. I’ll also include a short script you can run to verify basic GPU isolation on a Linux host. If you want that checklist and the verification script, I’ll produce them now.