☁️ Debugging clouds and cracking puns💻 #SatireInProgress
London
Joined February 2021
- Tweets2.1K
- Following181
- Followers17
- Likes6.8K
Amit Patel retweeted
There's also no reason to emulate massive organizations when you're small. The scale is off, the lessons are wrong, the insights don't fit. If you're interested in following a groove, find a fresh one that's closer to your size, not a huge one in dried mud.
OTD 11NOV1965 It was the last recording session for ‘Rubber Soul’. The songs ‘The Word’, Girl, ‘Wait’, I’m Looking Through You’ were recorded. The #Beatles also completed the recording of ‘You Won’t See Me’ with Mal Evans playing a one-note Hammond organ part. #RubberSoul
Amit Patel retweeted
github.com/jakubkrehel/skill… - Open-source AI interface design skills
rareui.com - 28+ free animated React components
reactbits.dev - Free animated React UI components
saurabhbhayana.com/component… - Free interactive React visual effects
skillsmd.dev - 91K+ AI agent skills indexed
🇪🇺 European gas is up again, and winter has barely started.
🔹 Dutch TTF: about €85 per MWh, 2.7 times a year ago
🔹 Peak: €103.56 in late October, the highest since 2022
🔹 EU storage: 74–75% full, against a 5-year average near 90%
🔹 🇩🇪 Germany: under 61%
Europe went into November with the least gas in storage since 2013, per the IEA.
Every cold week now comes straight out of storage.
I flagged this in my previous analysis: themerchantsnews.substack.co…
Commit conventions shouldn’t need another review comment.
Commit Check is a commit-policy validation tool for teams and AI agents that need consistent Git contributions.
It helps you catch policy violations before they merge by checking commit messages, branch names, and other Git metadata locally or in CI against a shared configuration.
Key features:
• Zero-config defaults – check Conventional Commits and Conventional Branch naming without writing a configuration file
• Shared policy file – use one cchk.toml across commit hooks, CI, and supported integrations
• More than messages – validate author names and emails, sign-offs, tag names, and committed file sizes or paths
• Pre-commit fixes – correct mechanical message mistakes, such as type capitalization or a missing colon, with --fix
• Agent-readable feedback – JSON output includes rule IDs, suggestions, and corrected values when a mechanical fix is available
It’s open-source (MIT license).
Link in the reply 👇
Amit Patel retweeted
🇲🇿 Mozambique has $ 27bn of LNG under construction, with up to $ 30bn more waiting for a signature.
🔹 Mozambique LNG (TotalEnergies): $ 20bn, 13m tonnes a year, built by Saipem, Chiyoda and McDermott. First LNG targeted for 2029.
🔹 Coral North FLNG (Eni): $ 7.2bn, 3.6m tonnes a year, built by Technip Energies, JGC and Samsung Heavy. Start-up targeted for 2028.
🔹 Rovuma LNG (ExxonMobil): 18.6m tonnes a year, FID targeted for the end of 2026, with letters of intent to Saipem, McDermott, Daewoo E&C, CPECC and Jan De Nul.
Up to $ 57bn of contracts, and the builders get paid first.
One name shows up in all 3: Saipem.
That’s where the oil money goes: themerchantsnews.substack.co…
🚨Eni is leaving Rovuma LNG, the giant onshore export project in 🇲🇿 Mozambique, through an asset swap with ExxonMobil.
🔹 ExxonMobil doubles its stake from 25% to 50% and stays operator
🔹 Eni takes other assets in the region and backs floating LNG
🔹 Rovuma LNG: 18.6m tonnes a year, $27–30bn, first LNG targeted for 2031
🔹 Final investment decision targeted by the end of 2026, with $ 1.1bn already committed to long-lead equipment
The partners left with ExxonMobil are CNPC, KOGAS, XRG and ENH.
Both companies are drawing on the same gas: Area 4 in the Rovuma Basin, roughly 85 trillion cubic feet.
ExxonMobil is betting on scale.
One onshore plant, 18.6m tonnes a year, built to run for decades.
Eni is betting on speed.
Coral South has been floating offshore and shipping cargoes since 2022.
Coral Norte, a $ 7.2bn 2nd unit, is under construction.
Floating LNG is smaller, but it comes online faster and stays offshore, away from the insurgency in Cabo Delgado that froze onshore LNG in the province for years.
Africa’s biggest undeveloped gas province will now be run 2 ways at once.
In a year when Qatar’s LNG was trapped behind Hormuz, buyers will pay for gas that sails straight into the Indian Ocean.
Whoever delivers 1st wins the contracts.
Amit Patel retweeted
Stop guessing what your coding agent remembers.
MemoSync is a local memory workbench for builders using Claude Code, Codex, Pi, or OpenCode across coding sessions.
It helps you inspect and steer agent memory by turning remembered rules into reviewable items, letting you choose what enters each turn, and checking how those items affected the result.
Key features:
• Reviewable memory changes – accept, edit, rescope, or dismiss proposed lessons before they reach the store in review mode.
• Per-turn memory selection – see which items the agent plans to use, add or remove them, then confirm before it starts.
• Memory citations – inspect the memories referenced in replies and stop a turn to correct how one is applied.
• Post-turn audit – a separate model pass checks selected memories for violations or influence; its judgments can be wrong.
• Memory Board – browse personal, project, and session memories, edit their scope, and inspect version history.
Free public GitHub repo.
Link in the reply 👇
Amit Patel retweeted
Sending a large file seems simple at first.
But there are many processes working together in the background.
In this tutorial, @reetain_raina covers chunk management, object storage, encryption, & everything else needed to upload a large file & share a download link.
freecodecamp.org/news/how-fi…
Amit Patel retweeted
It's even more funnier how the colonel embodies this duality himself by ordering the Joker to kill the enemies of America while also talking about making peace in Vietnam and it's terrifying how he doesn't seem aware of his contradiction
Amit Patel retweeted
estorides — Aggregates 99+ free OSINT sources in parallel, resolves entities into a knowledge graph, and exports STIX 2.1, MISP, or GraphML with LLM-assisted analysis. ktp.sh/JxIkFtGl5z
Amit Patel retweeted
🚨 Indian-origin Anil Chakravarthy is appointed as Adobe’s next president and CEO, from Dec 1, 2026.
Amit Patel retweeted
Testify, sister! I spent a lot of my time, from the Start Menu to Pinball to Task Manager, on things where latency sucks... because I hate latency.
One of my core tenets for the first Media Center prototype was that an animated piece of UI did NOT have to finish animating before it was active. So if the start menu flew in from the left on your TV, it was active from the first pixel being shown, not once it landed. To me, it made a huge difference in usability.
I can't even play certain slot machines if there's a 25ms delay between pressing SPIN and the reels moving. Drives me crazy!
There's only one real counter-example I can think of: the 400ms delay on Windows menu flyouts, so you've got a chance for your aim to settle before a submenu does the work of populating.
UX latency is the most frustrating (lack of) advancement in computing over the last ~50 years.
IBM, way back in the 80s, found that user productivity shot up *exponentially* as system response time dropped. The "Doherty Threshold" was defined as ~400 milliseconds; the maximum amount of tolerable UX latency before interrupting user flow state.
It's unbelievable how many modern operating systems (*cough* win11) violate this threshold constantly.
Literally all you need to do as a Dev is pay attention. Add a "Doherty Threshold" to your test suite. There's so many different profiling tools, and especially with how good models are, really no excuse. Sure, some stalls are unavoidable, but that's when you pull out other UX tricks to at least give the illusion of responsiveness.
Crazy how much vintage computing hardware feels immeasurably more responsive despite having thousands of times less overall performance.
Amit Patel retweeted
OTD 10NOV1961 The #Beatles performed at the Tower Ballroom, New Brighton, as part of the Big-Beat operation. This was a festival with Gerry and the Pacemakers, Rory Storm and the Hurricanes, Remo Four, and Kingsize Taylor and the Dominoes. #TheBeatles
Amit Patel retweeted
DEVELOPING: A new trailer for Jesse Eisenberg’s The Debut just released & I cannot wait to see this film. It’s time for all of us to lock in for the next 6 weeks as we wrap up 2026. A handful of major releases are on the way still & this is at the top of the list. Featuring Julianne Moore & Paul Giamatti, it’s not that often we get a musical-comedy. I’m excited to see what Eisenberg has in store for us.
Amit Patel retweeted
Every child, no matter who they are or where they live in the world, deserves the opportunity to walk, run, play and thrive.”
Full interview👉bit.ly/4j6EDM5
3 Questions, 2 Drops, 1 Goal: #EndPolio
@WHO_Polio @UNICEFPolio @EndPolioNow @GatesFoundation @CDCgov @Gavi @WHOAfro @UNICEFAfrica @WHOEmro @GatesAfrica @JGuudle @Rotary @ProfJanabi @HananBalkhy @DrGediMohamed @UNICEFPolioDir @KSRelief_En @EIB @CanadaDev @MBZFoundation @EU_Commission @UNFoundation @ChrisJElias @UNICEFChief @SaniaNishtar
Amit Patel retweeted
Open source heartbeat monitoring for cron jobs, workers and scheduled tasks that alerts you when a job goes quiet, self-hosted with Docker or hosted
ALT SilenceWatch promotional card: open-source heartbeat monitoring for cron jobs and workers that sends alerts when a job stops checking in. Tags include Freemium (Subscription), Open Source (Apache-2.0), Online, SaaS, Self-Hosted, Docker, France, and EU. Dark dashboard shows checks with status badges (Down, Late, Up), environments, schedules, and last ping times. Footer reads: Alternative to Uptime Kuma and Better Stack.
Amit Patel retweeted
GPU scheduling in Kubernetes is becoming an important skill now.
Here is why 👇
Most organizations are either experimenting or running ML workloads in Kubernetes.
As Devops engineers you should know how to set up and manage GPU resources.
We have created a simple guide on setting up GPU node using NVIDIA GPU Operator
By the end of this blog, you will have a clear understanding of:
- Need for GPU operator on Kubernetes
- Setting up NVIDIA GPU Operator on a Kubernetes cluster
- Verify if Kubernetes detects GPUs
- How to deploy a real GPU-based workload to validate the full stack.
To be honest, it is not as complicated as it sounds once you understand the basics.
Even if you dont have access to GPU's, just go through the content to have an understanding on how it works.
𝗥𝗲𝗮𝗱 𝗶𝘁 𝗵𝗲𝗿𝗲: devopscube.com/setup-gpu-ope…
#devops #mlops
Amit Patel retweeted
I've got local AI working in a loop on my dual RTX6000s in my @Dell7975 Workstation.
I'm letting it experiment to see if the GLM5.3-Flash AI can improve my custom Robotron-playing AI. I'm having it create a hypothesis, A/B test it, and pick the winner as the new control, then repeat the loop forever.
Here's the prompt. I'm no expert; I'd be curious about how to improve it!
-----
OK, I want you to continue the Robotron A/B hyperparameter testing. Your instructions are to:
1) Analyze the Robotron project, model, structure, and architecture, looking for opportunities to significantly improve learning performance without adversely impacting performance without merit (ie: it can be a little slower if much better, but not much slower for a little better!). Create a hypothesis for an improvement that can be A/B tested.
2) Then A/B test the hypothesis for up to 30 minutes. Each A/B test gets 96 games, so initially 192 clients run, all unthrottled. When a game finishes, its client process exits after its score is recorded so that remaining games run faster. If a game does not finish in the 30 minutes, but is still playing, then its current score is recorded, and the client ends. So at the end of 30 minutes, each client game will have a score.
3) Average the scores, looking for the highest performer. Compare your A/B clients scores to see if the test revealed an improvement. That improvement should be statistically valid, at least 5%, 99 times out of 100 (confidence interval).
3a) Report the results, indicating what you were testing, what the result was, and so on. Keep a running log called results.md in the root that I can check to see what you are currently doing and the results so far.
4) Pick the winner (control or new) as the new control, then go back to step 1 to repeat this process of improvement testing indefinitely
