Keyword matches words; semantic matches meaning and exact titles or names; hybrid combines both. In keyword mode, use quotation marks for exact phrases. Author matches published credits. Year applies to posts; wiki entries have no author or publication year. You can also use author:"Ada" or type:podcast in non-regex searches.
author:"Ada"
type:podcast
Explore articles, conversations, and the wiki.
Articles, conversations, profiles, and grounded writeups.
Select Apply filters to update results.
Search or filters have changed. The results below still show the previous search; select Search or Apply filters to update them.
721 results · Filters applied
…there is very real work to upgrade IT systems, get agents the context they need, modernize the workflows to work with agents, figure out the human-agent relationship in the workflow, drive adoption and do change management, and much more.…
…multiple languages. This limitation is highlighted in the context of handling Chinese and other languages, with users expressing the need for models to handle multilingual tasks effectively. There are concerns about the instruction template…
…persona behaviors (nearcyan, NickEMoran question). Practical engineering consequence: Anthropic also posted about an internal performance engineering take-home that became solvable by Opus 4.5, forcing them to redesign hiring evaluation—a…
…3 Opus, best of 4 runs. We are working on clustering and flow engineering with Haiku. LLMs and AI Models Llama 3 Performance: @abacaj noted that llama-3 models with zero-training can get 32k context with exceptional quality, surpassing…
…platform algorithms: X “For You” recommender goes public X Engineering open-sources the X algorithm (Grok-style transformer recommender): X says it has open-sourced its new algorithm (the ranking/recommendation stack), “powered by the same …
…than in November 2025. Customer Support rose 32 times and Engineering rose 27 times, while Legal grew more gradually but still reached 13 times its November level. This should form an interesting baseline against Tokenmaxxing concerns -…
…React Python, with Charuru emphasizing Sonnet's unique prompt engineering that enhances its human-like interactions. Open Source and Accessibility: DeepSeek V3 is celebrated for being open-source and accessible, allowing users to leverage…
…for Coding Benchmarks: Concerns arose about heavy prompt engineering in coding benchmarks. Some argued that simple prompts better reflect practical usage if the code fits within a reasonable context window, differing from agent evaluation…
…model inputs; @natefyi_30842 shared a link to a tokenizer for context. Interpretability and engineering representations: Max_paperclips discussed the exciting field of representations engineering, citing a favorite post and referring to…
…than only over batch or layers. Hugging Face docs at Accelerate context parallelism were recommended as a practical guide for combining these techniques. 4. Agent Frameworks, Tools, and Prompt/Behavior Engineering MCP Apps SDK Lets…
…centered on what he calls the “2026 metamorphosis” of software engineering. The notable signal is not just the course itself, but the curriculum reset: 85% of Fall 2025 material is being replaced with topics like agent skills, context…
…Heights: @gabriel_syme highlighted a repository focused on data engineering for scaling language models to 128K context, a significant advancement in the field. The VRAM requirements for such models at 7B scale exceed 600GB, a substantial…
Sign in with Google, or get the newsletter by email.
Loading Google sign-in…
Includes all newsletters: Latent Space (2–3 times a week), AI for Science (2–3 times a month), and AINews (every weekday). Free. Unsubscribe any time.
Uncheck to create an account without newsletter emails.
Discover technical newsletters we recommend, then explore Latent Space membership. Both are optional.
Your newsletter choice is optional. How we use your account data.
Please enable JavaScript to sign in or subscribe. You can still read the articles.