DeepSeek R1, the new entrant to the Large Language Model wars has created quite a splash over the last few weeks. Its entrance into a space dominated by the Big Corps, while pursuing asymmetric and novel strategies has been a refreshing eye-opener.GPT AI improvement was starting to show signs of slowing down, and has been observed to be reaching a point of diminishing returns as it runs out of data and compute required to train, fine-tune increasingly large models. This has turned the focus towards building "reasoning" models that are post-trained through reinforcement learning, techniques such as inference-time and test-time scaling and search algorithms to make the models appear to think and reason better. OpenAI&aposs o1-series models were the first to achieve this successfully with its inference-time scaling and Chain-of-Thought reasoning.
Intelligence as an emergent property of Reinforcement Learning (RL)Reinforcement Learning (RL) has been successfully used in the past by Google&aposs DeepMind team to build highly intelligent and specialized systems where intelligence is observed as an emergent property through rewards-based training approach that yielded achievements like AlphaGo (see my post on it here - AlphaGo: a journey to machine intuition).DeepMind went on to build a series of Alpha* projects that achieved many notable feats using RL:
AlphaGo, defeated the world champion Lee Seedol in the game of Go
AlphaZero, a generalized system that learned to play games such as Chess, Shogi and Go without human input
AlphaStar, achieved high performance in the complex real-time strategy game StarCraft II.
AlphaFold, a tool for predicting protein structures which significantly advanced computational biology.
AlphaCode, a model designed to generate computer programs, performing competitively in coding challenges.
AlphaDev, a system developed to discover novel algorithms, notably optimizing sorting algorithms beyond human-derived methods.
All of these systems achieved mastery in its own area through self-training/self-play and by optimizing and maximizing the cumulative reward over time by interacting with its environment where intelligence was observed as an emergent property of the system.The RL feedback loopRL mimics the process through which a baby would learn to walk, through trial, error and first principles.
R1 model training pipelineAt a technical level, DeepSeek-R1 leverages a combination of Reinforcement Learning (RL) and Supervised Fine-Tuning (SFT) for its training pipeline:DeepSeek-R1 Model Training PipelineUsing RL and DeepSeek-v3, an interim reasoning model was built, called DeepSeek-R1-Zero, purely based on RL without relying on SFT, which demonstrated superior reasoning capabilities that matched the performance of OpenAI&aposs o1 in certain benchmarks such as AIME 2024.The model was however affected by poor readability and language-mixing and is only an interim-reasoning model built on RL principles and self-evolution.DeepSeek-R1-Zero was then used to generate SFT data, which was combined with supervised data from DeepSeek-v3 to re-train the DeepSeek-v3-Base model.The new DeepSeek-v3-Base model then underwent additional RL with prompts and scenarios to come up with the DeepSeek-R1 model.The R1-model was then used to distill a number of smaller open source models such as Llama-8b, Qwen-7b, 14b which outperformed bigger models by a large margin, effectively making the smaller models more accessible and usable.
Key contributions of DeepSeek-R1
RL without the need for SFT for emergent reasoning capabilities
R1 was the first open research project to validate the efficacy of RL directly on the base model without relying on SFT as a first step, which resulted in the model developing advanced reasoning capabilities purely through self-reflection and self-verification.Although, it did degrade in its language capabilities during the process, its Chain-of-Thought (CoT) capabilities for solving complex problems was later used for further RL on the DeepSeek-v3-Base model which became R1. This is a significant contribution back to the research community.The below analysis of DeepSeek-R1-Zero and OpenAI o1-0912 shows that it is viable to attain robust reasoning capabilities purely through RL alone, which can be further augmented with other techniques to deliver even better reasoning performance.Source: https://github.com/deepseek-ai/DeepSeek-R1Its quite interesting, that the application of RL gives rise to seemingly human capabilities of "reflection", and arriving at "aha" moments, causing it to pause, ponder and focus on a specific aspect of the problem, resulting in emergent capabilities to problem-solve as humans do.
Model distillation
DeepSeek-R1 also demonstrated that larger models can be distilled into smaller models which makes advanced capabilities accessible to resource-constrained environments, such as your laptop. While its not possible to run a 671b model on a stock laptop, you can still run a distilled 14b model that is distilled from the larger model which still performs better than most publicly available models out there. This enables intelligence to be brought closer to the edge, to allow faster inference at the point of experience (such as on a smartphone, or on a Raspberry Pi), which paves way for more use cases and possibilities for innovation.Source: https://github.com/deepseek-ai/DeepSeek-R1Distilled models are very different to R1, which is a massive model with a completely different model architecture than the distilled variants, and so are not directly comparable in terms of capability, but are instead built to be more smaller and efficient for more constrained environments. This technique of being able to distill a larger model&aposs capabilities down to a smaller model for portability, accessibility, speed, and cost will bring about a lot of possibilities for applying artificial intelligence in places where it would have otherwise not been possible. This is another key contribution of this technology from DeepSeek, which I believe has even further potential for democratization and accessibility of AI.
Why is this moment so significant?DeepSeek-R1 was a pivotal contribution in many ways.
The contributions to the state-of-the-art and the open research helps move the field forward where everybody benefits, not just a few highly funded AI labs building the next billion dollar model.
Open-sourcing and making the model freely available follows an asymmetric strategy to the prevailing closed nature of much of the model-sphere of the larger players. DeepSeek should be commended for making their contributions free and open.
It reminds us that its not just a one-horse race, and it incentivizes competition, which has already resulted in OpenAI o3-mini a cost-effective reasoning model which now shows the Chain-of-Thought reasoning. Competition is a good thing.
We stand at the cusp of an explosion of small-models that are hyper-specialized, and optimized for a specific use case that can be trained and deployed cheaply for solving problems at the edge. It raises a lot of exciting possibilities and is why DeepSeek-R1 is one of the most pivotal moments of tech history.
Twenty years ago, it was easy to dislike Microsoft. It was the quintessential evil MegaCorp that was quick to squash competition, often ruthlessly, but in some cases slowly through a more insidious process of embracing, extending, and exterminating anything that got in the way. This was the signature personality of Ballmer-era Microsoft that also inspired and united the software freedom fighting forces that came together to safeguard things that mattered to them and were at risk.I remember the era when the Novell, SCO, and Microsoft saga cast fear, uncertainty, and doubt on the future of open Unix and Linux and on what would happen to the operating systems that we loved if the suits of Redmond prevailed. Looking back, I&aposm glad that the arc of this story has bent towards justice, and I shudder at the possibilities had it worked out differently.Looking at today&aposs Microsoft, I&aposm amazed at how much change a leader with the right vision can make to the trajectory of a company that even makes an old-school software freedom advocate as me admire and even applaud the strides it has taken in the last 10 or so years that has dramatically shifted the perception of Microsoft. The personality of the Satya-era Microsoft is one to behold. While it will take more time to win back the trust, we see the tides changing and the positivity is important for the entire industry.For Microsoft, it was TypeScript and VS Code that helped change the narrative internally which led to its internal resurgence and acceptance of open source. Its acquisition of GitHub propelled it forward within the community overnight. Its contributions to the Linux kernel and other major software projects have also been consequential in changing its public perceptions.It takes a while to claw back trust and is very easy to breach. This time, however, Microsoft seems to understand this dynamic more than it did 20 years ago. All it took was the right leadership.
US-11604662-B2I m happy to announce, that after a long wait, patent US-11604662-B2 has been issued.I want to thank and recognize my co-inventors, Div Prakash and Subin George, who I m privileged to be on paper with.The effort that led to this work involved a group of engineers, many phone calls, some nerve-wracking presentations, culminating in a fantastic hackathon-winning outcome for a young and talented team, which I was proud to be a part of and privileged to lead.
Parallelizing the compilation of a large codebase is a breeze with distcc, which allows you to spread the load across multiple nodes and speed up the compilation time.Here s a sample network topology for a distributed build:Install distcc on the three Debian/Ubuntu-based nodes:
# apt install distcc
Edit /etc/default/distcc and set:
STARTDISTCC="true"
# Customize for your environment ALLOWEDNETS="192.168.2.0/24"
# Specify your network device LISTENER="192.168.2.146"
Additionally, the JOBS and NICE variables can be tweaked to suit the compute power that you have available.Start distcc:
# systemctl start distcc
Do the same all the nodes, and if you have a firewall enabled with ufw, you will need to open up the port 3632 to the master node.
# ufw allow 3632/tcp
Additionally, if you d like to use ssh over untrusted networks so code and communication with the worker nodes happen over a secure channel, ensure that SSH is running and is opened up to the master node in the same manner as above with the key of the master node in ~/.ssh/authorized_keys of the worker nodes. Opening port 3632 in this manner is a security hole, so take precautions over untrusted networks.Back in the master node, setup a DISTCC_HOSTS environment variable that lists the worker nodes, including the master node. Note the order of the hosts, as it is important. The first host will be more heavily used, and distcc has no way of knowing the capacity and capability of the hosts, so specify the most powerful host first.
At this point, you re ready to compile.Go to your codebase, in this case we use the Linux kernel source code for the purpose of example.
$ make tinyconfig $ time make -j$(nproc) CC=distcc bzImage
On another terminal, you can monitor the status of the distributed compilation with distmoncc-text or tools such as top or bpytop.Network throughput and latency will be a big factor in how much distcc will help speed up your build process. Using ssh may additionally introduce overhead, so play with the variables to see how much distcc can help speed up or optimize the build for your specific scenario. You may want to additionally consider ccache to speed up the build process.There are some aspects of the build process that are not effectively parallizable in this manner, such as the final linking step of the executable, for which you will not see any performance improvement with distcc.Give distcc a spin, and put any spare compute you have lying around in your home lab to good use.
If you ve done anything in the Kubernetes space in recent years, you ve most likely come across the words Service Mesh . It s backed by a set of mature technologies that provides cross-cutting networking, security, infrastructure capabilities to be used by workloads running in Kubernetes in a manner that is transparent to the actual workload. This abstraction enables application developers to not worry about building in otherwise sophisticated capabilities for networking, routing, circuit-breaking and security, and simply rely on the services offered by the service mesh.In this post, I ll be covering Linkerd, which is an alternative to Istio. It has gone through a significant re-write when it transitioned from the JVM to a Go-based Control Plane and a Rust-based Data Plane a few years back and is now a part of the CNCF and is backed by Buoyant. It has proven itself widely for use in production workloads and has a healthy community and release cadence.It achieves this with a side-car container that communicates with a Linkerd control plane that allows central management of policy, telemetry, mutual TLS, traffic routing, shaping, retries, load balancing, circuit-breaking and other cross-cutting concerns before the traffic hits the container. This has made the task of implementing the application services much simpler as it is managed by container orchestrator and service mesh. I covered Istio in a prior post a few years back, and much of the content is still applicable for this post, if you d like to have a look.Here are the broad architectural components of Linkerd:The components are separated into the control plane and the data plane.The control plane components live in its own namespace and consists of a controller that the Linkerd CLI interacts with via the Kubernetes API. The destination service is used for service discovery, TLS identity, policy on access control for inter-service communication and service profile information on routing, retries, timeouts. The identity service acts as the Certificate Authority which responds to Certificate Signing Requests (CSRs) from proxies for initialization and for service-to-service encrypted traffic. The proxy injector is an admission webhook that injects the Linkerd proxy side car and the init container automatically into a pod when the linkerd.io/inject: enabled is available on the namespace or workload.On the data plane side are two components. First, the init container, which is responsible for automatically forwarding incoming and outgoing traffic through the Linkerd proxy via iptables rules. Second, the Linkerd proxy, which is a lightweight micro-proxy written in Rust, is the data plane itself.I will be walking you through the setup of Linkerd (2.12.2 at the time of writing) on a Kubernetes cluster.Let s see what s running on the cluster currently. This assumes you have a cluster running and kubectl is installed and available on the PATH.
On most systems, this should be sufficient to setup the CLI. You may need to restart your terminal to load the updated paths. If you have a non-standard configuration and linkerd is not found after the installation, add the following to your PATH to be able to find the cli:
export PATH=$PATH:~/.linkerd2/bin/
At this point, checking the version would give you the following:
$ linkerd version Client version: stable-2.12.2 Server version: unavailable
Setting up Linkerd Control PlaneBefore installing Linkerd on the cluster, run the following step to check the cluster for pre-requisites:
kubernetes-api -------------- can initialize the client can query the Kubernetes API
kubernetes-version ------------------ is running the minimum Kubernetes API version is running the minimum kubectl version
pre-kubernetes-setup -------------------- control plane namespace does not already exist can create non-namespaced resources can create ServiceAccounts can create Services can create Deployments can create CronJobs can create ConfigMaps can create Secrets can read Secrets can read extension-apiserver-authentication configmap no clock skew detected
linkerd-version --------------- can determine the latest version cli is up-to-date
Status check results are
All the pre-requisites appear to be good right now, and so installation can proceed.The first step of the installation is to setup the Custom Resource Definitions (CRDs) that Linkerd requires. The linkerd cli only prints the resource YAMLs to standard output and does not create them directly in Kubernetes, so you would need to pipe the output to kubectl apply to create the resources in the cluster that you re working with.
$ linkerd install --crds kubectl apply -f - Rendering Linkerd CRDs... Next, run linkerd install kubectl apply -f - to install the control plane.
customresourcedefinition.apiextensions.k8s.io/authorizationpolicies.policy.linkerd.io created customresourcedefinition.apiextensions.k8s.io/httproutes.policy.linkerd.io created customresourcedefinition.apiextensions.k8s.io/meshtlsauthentications.policy.linkerd.io created customresourcedefinition.apiextensions.k8s.io/networkauthentications.policy.linkerd.io created customresourcedefinition.apiextensions.k8s.io/serverauthorizations.policy.linkerd.io created customresourcedefinition.apiextensions.k8s.io/servers.policy.linkerd.io created customresourcedefinition.apiextensions.k8s.io/serviceprofiles.linkerd.io created
Next, install the Linkerd control plane components in the same manner, this time without the crds switch:
$ linkerd install kubectl apply -f - namespace/linkerd created clusterrole.rbac.authorization.k8s.io/linkerd-linkerd-identity created clusterrolebinding.rbac.authorization.k8s.io/linkerd-linkerd-identity created serviceaccount/linkerd-identity created clusterrole.rbac.authorization.k8s.io/linkerd-linkerd-destination created clusterrolebinding.rbac.authorization.k8s.io/linkerd-linkerd-destination created serviceaccount/linkerd-destination created secret/linkerd-sp-validator-k8s-tls created validatingwebhookconfiguration.admissionregistration.k8s.io/linkerd-sp-validator-webhook-config created secret/linkerd-policy-validator-k8s-tls created validatingwebhookconfiguration.admissionregistration.k8s.io/linkerd-policy-validator-webhook-config created clusterrole.rbac.authorization.k8s.io/linkerd-policy created clusterrolebinding.rbac.authorization.k8s.io/linkerd-destination-policy created role.rbac.authorization.k8s.io/linkerd-heartbeat created rolebinding.rbac.authorization.k8s.io/linkerd-heartbeat created clusterrole.rbac.authorization.k8s.io/linkerd-heartbeat created clusterrolebinding.rbac.authorization.k8s.io/linkerd-heartbeat created serviceaccount/linkerd-heartbeat created clusterrole.rbac.authorization.k8s.io/linkerd-linkerd-proxy-injector created clusterrolebinding.rbac.authorization.k8s.io/linkerd-linkerd-proxy-injector created serviceaccount/linkerd-proxy-injector created secret/linkerd-proxy-injector-k8s-tls created mutatingwebhookconfiguration.admissionregistration.k8s.io/linkerd-proxy-injector-webhook-config created configmap/linkerd-config created secret/linkerd-identity-issuer created configmap/linkerd-identity-trust-roots created service/linkerd-identity created service/linkerd-identity-headless created deployment.apps/linkerd-identity created service/linkerd-dst created service/linkerd-dst-headless created service/linkerd-sp-validator created service/linkerd-policy created service/linkerd-policy-validator created deployment.apps/linkerd-destination created cronjob.batch/linkerd-heartbeat created deployment.apps/linkerd-proxy-injector created service/linkerd-proxy-injector created secret/linkerd-config-overrides created
Kubernetes will start spinning up the data plane components and you should see the following when you list the pods:
kubernetes-api -------------- can initialize the client can query the Kubernetes API
kubernetes-version ------------------ is running the minimum Kubernetes API version is running the minimum kubectl version
linkerd-existence ----------------- 'linkerd-config' config map exists heartbeat ServiceAccount exist control plane replica sets are ready no unschedulable pods control plane pods are ready cluster networks contains all pods cluster networks contains all services
linkerd-config -------------- control plane Namespace exists control plane ClusterRoles exist control plane ClusterRoleBindings exist control plane ServiceAccounts exist control plane CustomResourceDefinitions exist control plane MutatingWebhookConfigurations exist control plane ValidatingWebhookConfigurations exist proxy-init container runs as root user if docker container runtime is used
linkerd-identity ---------------- certificate config is valid trust anchors are using supported crypto algorithm trust anchors are within their validity period trust anchors are valid for at least 60 days issuer cert is using supported crypto algorithm issuer cert is within its validity period issuer cert is valid for at least 60 days issuer cert is issued by the trust anchor
linkerd-webhooks-and-apisvc-tls ------------------------------- proxy-injector webhook has valid cert proxy-injector cert is valid for at least 60 days sp-validator webhook has valid cert sp-validator cert is valid for at least 60 days policy-validator webhook has valid cert policy-validator cert is valid for at least 60 days
linkerd-version --------------- can determine the latest version cli is up-to-date
control-plane-version --------------------- can retrieve the control plane version control plane is up-to-date control plane and cli versions match
linkerd-control-plane-proxy --------------------------- control plane proxies are healthy control plane proxies are up-to-date control plane proxies and cli versions match
Status check results are
Everything looks good.Setting up the Viz ExtensionAt this point, the required components for the service mesh are setup, but let s also install the viz extension, which provides a good visualization capabilities that will come in handy subsequently. Once again, linkerd uses the same pattern for installing the extension.
$ linkerd viz install kubectl apply -f - namespace/linkerd-viz created clusterrole.rbac.authorization.k8s.io/linkerd-linkerd-viz-metrics-api created clusterrolebinding.rbac.authorization.k8s.io/linkerd-linkerd-viz-metrics-api created serviceaccount/metrics-api created clusterrole.rbac.authorization.k8s.io/linkerd-linkerd-viz-prometheus created clusterrolebinding.rbac.authorization.k8s.io/linkerd-linkerd-viz-prometheus created serviceaccount/prometheus created clusterrole.rbac.authorization.k8s.io/linkerd-linkerd-viz-tap created clusterrole.rbac.authorization.k8s.io/linkerd-linkerd-viz-tap-admin created clusterrolebinding.rbac.authorization.k8s.io/linkerd-linkerd-viz-tap created clusterrolebinding.rbac.authorization.k8s.io/linkerd-linkerd-viz-tap-auth-delegator created serviceaccount/tap created rolebinding.rbac.authorization.k8s.io/linkerd-linkerd-viz-tap-auth-reader created secret/tap-k8s-tls created apiservice.apiregistration.k8s.io/v1alpha1.tap.linkerd.io created role.rbac.authorization.k8s.io/web created rolebinding.rbac.authorization.k8s.io/web created clusterrole.rbac.authorization.k8s.io/linkerd-linkerd-viz-web-check created clusterrolebinding.rbac.authorization.k8s.io/linkerd-linkerd-viz-web-check created clusterrolebinding.rbac.authorization.k8s.io/linkerd-linkerd-viz-web-admin created clusterrole.rbac.authorization.k8s.io/linkerd-linkerd-viz-web-api created clusterrolebinding.rbac.authorization.k8s.io/linkerd-linkerd-viz-web-api created serviceaccount/web created server.policy.linkerd.io/admin created authorizationpolicy.policy.linkerd.io/admin created networkauthentication.policy.linkerd.io/kubelet created server.policy.linkerd.io/proxy-admin created authorizationpolicy.policy.linkerd.io/proxy-admin created service/metrics-api created deployment.apps/metrics-api created server.policy.linkerd.io/metrics-api created authorizationpolicy.policy.linkerd.io/metrics-api created meshtlsauthentication.policy.linkerd.io/metrics-api-web created configmap/prometheus-config created service/prometheus created deployment.apps/prometheus created service/tap created deployment.apps/tap created server.policy.linkerd.io/tap-api created authorizationpolicy.policy.linkerd.io/tap created clusterrole.rbac.authorization.k8s.io/linkerd-tap-injector created clusterrolebinding.rbac.authorization.k8s.io/linkerd-tap-injector created serviceaccount/tap-injector created secret/tap-injector-k8s-tls created mutatingwebhookconfiguration.admissionregistration.k8s.io/linkerd-tap-injector-webhook-config created service/tap-injector created deployment.apps/tap-injector created server.policy.linkerd.io/tap-injector-webhook created authorizationpolicy.policy.linkerd.io/tap-injector created networkauthentication.policy.linkerd.io/kube-api-server created service/web created deployment.apps/web created serviceprofile.linkerd.io/metrics-api.linkerd-viz.svc.cluster.local created serviceprofile.linkerd.io/prometheus.linkerd-viz.svc.cluster.local created
A few seconds later, you should see the following in your pod list:
The viz components live in the linkerd-viz namespace.You can now checkout the viz dashboard:
$ linkerd viz dashboard Linkerd dashboard available at: http://localhost:50750 Grafana dashboard available at: http://localhost:50750/grafana Opening Linkerd dashboard in the default browser Opening in existing browser session.
The Meshed column indicates the workload that is currently integrated with the Linkerd control plane. As you can see, there are no application deployments right now that are running.Injecting the Linkerd Data Plane componentsThere are two ways to integrate Linkerd to the application containers:1 by manually injecting the Linkerd data plane components 2 by instructing Kubernetes to automatically inject the data plane componentsInject Linkerd data plane manuallyLet s try the first option. Below is a simple nginx-app that I will deploy into the cluster:
Back in the viz dashboard, I do see the workload deployed, but it isn t currently communicating with the Linkerd control plane, and so doesn t show any metrics, and the Meshed count is 0:Looking at the Pod s deployment YAML, I can see that it only includes the nginx container:
Let s directly inject the linkerd data plane into this running container. We do this by retrieving the YAML of the deployment, piping it to linkerd cli to inject the necessary components and then piping to kubectl apply the changed resources.
Back in the viz dashboard, the workload now is integrated into Linkerd control plane.Looking at the updated Pod definition, we see a number of changes that the linkerd has injected that allows it to integrate with the control plane. Let s have a look:
At this point, the necessary components are setup for you to explore Linkerd further. You can also try out the jaeger and multicluster extensions, similar to the process of installing and using the viz extension and try out their capabilities.Inject Linkerd data plane automaticallyIn this approach, we shall we how to instruct Kubernetes to automatically inject the Linkerd data plane to workloads at deployment time.We can achieve this by adding the linkerd.io/inject annotation to the deployment descriptor which causes the proxy injector admission hook to execute and inject linkerd data plane components automatically at the time of deployment.
This annotation can also be specified at the namespace level to affect all the workloads within the namespace. Note that any resources created before the annotation was added to the namespace will require a rollout restart to trigger the injection of the Linkerd components.Uninstalling LinkerdNow that we have walked through the installation and setup process of Linkerd, let s also cover how to remove it from the infrastructure and go back to the state prior to its installation.The first step would be to remove extensions, such as viz.
DeepSeek R1, the new entrant to the Large Language Model wars has created quite a splash over the last few weeks. Its entrance into a space dominated by the Big Corps, while pursuing asymmetric and novel strategies has been a refreshing eye-opener.GPT AI improvement was starting to show signs of slowing down, and has been observed to be reaching a point of diminishing returns as it runs out of data and compute required to train, fine-tune increasingly large models. This has turned the focus towards building "reasoning" models that are post-trained through reinforcement learning, techniques such as inference-time and test-time scaling and search algorithms to make the models appear to think and reason better. OpenAI&aposs o1-series models were the first to achieve this successfully with its inference-time scaling and Chain-of-Thought reasoning.
The RL feedback loop
DeepSeek-R1 Model Training Pipeline
Source: https://github.com/deepseek-ai/DeepSeek-R1
Source: https://github.com/deepseek-ai/DeepSeek-R1
Twenty years ago, it was easy to dislike Microsoft. It was the quintessential evil MegaCorp that was quick to squash competition, often ruthlessly, but in some cases slowly through a more insidious process of embracing, extending, and exterminating anything that got in the way. This was the signature personality of Ballmer-era Microsoft that also inspired and united the software freedom fighting forces that came together to safeguard things that mattered to them and were at risk.I remember the era when the Novell, SCO, and Microsoft saga cast fear, uncertainty, and doubt on the future of open Unix and Linux and on what would happen to the operating systems that we loved if the suits of Redmond prevailed. Looking back, I&aposm glad that the arc of this story has bent towards justice, and I shudder at the possibilities had it worked out differently.Looking at today&aposs Microsoft, I&aposm amazed at how much change a leader with the right vision can make to the trajectory of a company that even makes an old-school software freedom advocate as me admire and even applaud the strides it has taken in the last 10 or so years that has dramatically shifted the perception of Microsoft. The personality of the Satya-era Microsoft is one to behold. While it will take more time to win back the trust, we see the tides changing and the positivity is important for the entire industry.For Microsoft, it was TypeScript and VS Code that helped change the narrative internally which led to its internal resurgence and acceptance of open source. Its acquisition of GitHub propelled it forward within the community overnight. Its contributions to the Linux kernel and other major software projects have also been consequential in changing its public perceptions.It takes a while to claw back trust and is very easy to breach. This time, however, Microsoft seems to understand this dynamic more than it did 20 years ago. All it took was the right leadership.
Testing 1 2 3
US-11604662-B2
The following contributors got their Debian Developer accounts in the last two months: