When you have a problem, just code a solution. I suck at time tracking and management. I tried apps like ActivityWatch that just didn't work well, and apps like RIZE charge $13 a month. So I built this overnight. Local Mac native time management tool.

Dear Desi parents, I get it. You love your kids and want the best for them. But please stop applying for internships and placements on their behalf. It doesn’t look good to employers. It signals that the candidate can’t (or won’t) do the basic work themselves. Let them apply. Let them own it. That’s part of growing up. I didn’t want to say this, but it’s happening way too often now and it’s a real put-off. You are actually doing a disservice to your kids in more ways than one.

So single that I had to go out with @_notapenguin today 😭

Anyone cosplaying as a data Center this Halloween??

i am speechless..

Everyone thinks they can build AI Governance in house. Yeah sure why don't you go ahead and build: - Enterprise AI Asset Discovery Engine - Shadow AI Behavioral Detection - Unmanaged LLM Endpoint Discovery - AI Application Fingerprinting - Cross-Cloud AI Resource Enumeration - SaaS AI Feature Activation Detection - Local Model Runtime Identification - Enterprise AI Software Bill of Materials (AI-BOM) - Unregistered AI Agent Enumeration - AI Usage Attribution Mapping - AI Service Ownership Resolution - Model Provider Classification Engine - Unauthorized Browser AI Extension Discovery - AI API Key Usage Discovery - AI Data Flow Topology Mapping - AI Workload Environment Classification - AI Asset Risk Scoring - Unsanctioned Model Deployment Detection - AI Infrastructure Drift Detection - Third-Party AI Dependency Discovery - Enterprise AI Adoption Heatmapping - Dormant AI Asset Identification - AI Asset Lifecycle Tracking - Continuous AI Asset Reconciliation - Shadow AI Risk Exposure Dashboard - AI-Specific Role-Based Access Control (RBAC) - Attribute-Based AI Authorization (ABAC) - Context-Aware Policy Enforcement - Just-in-Time AI Privilege Elevation - Time-Bound AI Resource Entitlements - AI Identity Federation - SCIM-Based AI User Provisioning - Machine-to-Machine Identity Governance - AI Service Account Privilege Auditing - AI API Credential Lifecycle Management - Least-Privilege AI Access Analysis - Agent Identity Attestation - Cross-Tenant AI Access Isolation - AI Access Request Orchestration - Multi-Stage AI Access Approval Chains - Privileged AI Operation Monitoring - AI Permission Change Forensics - Sensitive Model Access Segmentation - AI Resource Access Recertification - Conflicting AI Permission Detection - AI Session Risk Assessment - Adaptive AI Access Revocation - Break-Glass AI Administration Controls - Identity-to-Inference Traceability - Zero-Trust AI Policy Decision Engine - AI Prompt Data Loss Prevention (DLP) - AI Response Data Loss Prevention - Sensitive Data Classification Pipeline - PII Detection and Automated Redaction - PHI Exposure Prevention - Enterprise Secret Leakage Detection - Source Data Entitlement Verification - Retrieval-Augmented Generation Access Filtering - Cross-Domain AI Data Boundary Enforcement - AI Data Residency Policy Enforcement - Prompt Payload Encryption Enforcement - AI Data Retention Policy Automation - AI Conversation Deletion Orchestration - Sensitive Context Window Inspection - Embedding Data Sensitivity Classification - Vector Database Access Auditing - Unauthorized Dataset Exfiltration Detection - AI Output Confidentiality Labeling - External Model Data Exposure Assessment - AI Training Data Consent Verification - Privacy-Preserving Prompt Transformation - AI Data Minimization Enforcement - Contextual Data Leakage Risk Scoring - AI Data Lineage Reconstruction - AI Privacy Incident Evidence Preservation - Multi-Provider AI Cost Aggregation - Request-Level Token Accounting - Hierarchical AI Cost Allocation - Per-Agent Token Budget Enforcement - Per-User AI Spending Quotas - Per-Workflow AI Cost Attribution - Real-Time AI Budget Consumption Monitoring - Predictive AI Spend Forecasting - AI Spending Anomaly Detection - Token Waste Pattern Identification - Prompt-to-Completion Cost Decomposition - Model Cost-Efficiency Benchmarking - Dynamic Model Cost Routing - Budget-Aware Inference Throttling - AI Spend Hard-Stop Enforcement - AI Budget Exception Approval Workflow - AI Cost Center Chargeback Automation - AI Cost Showback Reporting - Inference Unit Economics Analysis - Token-to-Business-Value Correlation - AI Subscription Utilization Assessment - Idle AI Infrastructure Cost Detection - GPU Inference Utilization Optimization - AI Cost Optimization Engine - Tokenomics Command Center - Enterprise Model Registry Governance - Model Provenance Verification - Model Version Integrity Monitoring - AI Decision Trace Reconstruction - Cognitive Decisioning Explanation Engine

Anthropic recorded an $80 billion net loss in the first half of 2026. I have a small suggestion to help them recover their losses. Claude Max costs $200. Just sell Claude Max to 400 million people. 400M × $200 = $80 billion. Easy.

The IPO paper is about to get 1000x more citations ;)

are you fucking kidding me? pikzels.kumar-coding.dev/

@TheDeverakonda delivers one of the best performances I've seen in a long time in #Ranabali Bone chilling story and amazing screen presence. The entire team did an amazing job on the film and the cast delivers a tremendous pain filled story.

If such public space like this exist in India we wouldn't have to spend 2k ₹ in cafe just to meet friend.

Do a kickflip!

After reading for 15 mins: I think Nick is operating in a world where interpretability was just about enumerating the neurons and superposition was an obstacle to that plan, so now he's happy because maybe superposition isn't real. But: 1) there are v few empirical results. Unfortunately interp has a long history of significantly overclaiming how interpretable their findings are (SAEs, early circuits having poor causal scrubbing scores etc) so I'm not updating much at all on the evidence he's already shown. it's really got a long history of people getting overexcited. So looking forward to seeing the paper with all the empirical evidence! 2) I think it's not that big news for me even if true because superposition in something like the average case wasn't my main crux to thinking interpretability is tractable. some cruxes this doesn't really deal with at all: - I think the correct 'circuit' is usually just an overwhelmingly large bag of random-seeming heuristics (a la Redwood Research’s interpretation of how an LLM predicts is vs was) and having it isn't very informative/doesn't help with much/doesn't lead to good predictions. - More generally, It seems possible to enumerate all the 'low level' features in a neural network, and still have a hard time answering 'high level' questions like: "Is the model misaligned? / What would it do in this new situation? / If it is misaligned what should we do about it?" there's an abstraction/emergence gap to cross from the features to the questions of interest, and crossing that gap is likely much harder than enumerating features. That said, if his claims hold up, I think it will be quite a big deal, and a useful precursor to tackling the thornier problems! There may also be some useful things to do without tackling the thorny problems: eg we'll be able to make use of masses of automated agents to scour the features and circuits to try and make sense of the mess, and maybe that’s good for something. although I suspect that even the most optimistic outcome here will not make alignment of superhuman AIs much easier (as some on twitter have claimed) because the neuron interpretability will not be adversarially robust, among other issues

interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works already-clean.goodfire.com/