Tithon Brings Persistent Jupyter Kernels and Live Output Streaming to VSCode

·1 views

A developer has released Tithon, a VSCode tool designed to address a common pain point for machine learning practitioners running long Jupyter Notebook sessions. The tool maintains persistent Jupyter kernel execution, meaning training loops and other processes survive interruptions such as network drops or unexpected machine shutdowns. It also streams live cell output directly within the IDE, removing the need to monitor external tools like Weights & Biases or log files. The project was shared on Hacker News as a community showcase, where it has received early feedback. The developer built Tithon after repeatedly losing kernel state and cell outputs during multi-day model training runs.

Read the full story at Hacker News

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

How Distributed Transactions Work in Microservices: 2PC, Saga, and Kafka Explained

When an order spans multiple microservices — such as OrderService, PaymentService, and InventoryService — a standard local ACID transaction is insufficient because each service operates on its own database. Two-Phase Commit (2PC) attempts to enforce atomicity across services via a coordinator, but risks performance bottlenecks due to resource locking. The Saga pattern offers an alternative by breaking the workflow into independent local transactions, using compensating actions to reverse completed steps if a later stage fails. Saga coordination can be implemented via choreography, where services react to events published on a message broker like Kafka, or via orchestration, where a central coordinator directs each step. Because compensations are new business transactions rather than true rollbacks, Sagas provide eventual consistency rather than the strong consistency of a single atomic transaction.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Wrestles With Sharing Achievements Without Appearing Egotistical

A developer on DEV Community has written about the internal conflict of wanting to share professional accomplishments while fearing it may come across as arrogant or hurtful to others. The author describes feeling trapped in a lose-lose situation where sharing achievements risks seeming boastful, but staying silent invites doubt about their abilities. Growing up in an environment where emotions were suppressed has made it harder for them to read how others respond to their posts. Despite wanting to share wins for both portfolio-building and mental health reasons, the author plans to reduce their posting frequency to once a month. They intend to remain active on the platform through commenting and engaging with others' content instead.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How Sharing Project Stats With Context Can Build a Loyal Social Media Following

Developers and creators building in public can grow social media audiences by regularly sharing project metrics alongside meaningful context, not just raw numbers. Explaining what a statistic means, what influenced it, and what will be tested next turns routine data into useful content for other builders. Consistency matters: using the same core metric across updates lets followers track progress over time, including flat periods and failed experiments. Practical details such as labeling chart axes clearly, defining terms like 'active users', and removing sensitive information before posting screenshots help maintain credibility. Ending each update with a specific question invites readers to contribute their own experience, making the content a starting point for conversation rather than a one-way broadcast.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How Browser Simulators Can Teach Database Scaling Better Than Textbooks

A software developer argues that hands-on simulation is more effective than reading for understanding database scaling concepts like indexing, caching, and replication. Three free, browser-based simulators allow users to observe how missing indexes cause slow queries, how cache eviction policies affect hit rates, and how read replicas introduce data consistency challenges. The indexing simulator demonstrates how a B-tree index reduces rows scanned during a query, while also illustrating the write overhead that makes blanket indexing impractical. The caching simulator visualizes the cache-aside pattern and lets users compare LRU, FIFO, and LFU eviction strategies against different access patterns. The tools are disclosed as built by the author and require no signup, serving as a practical supplement to standard system design interview preparation.

0 comments Read more at DEV Community