The Task at Hand
Welcome to our blog, where we share insights, updates, and stories about Taskless and what it takes to make AI systems more deterministic. Here you'll find a mix of technical deep-dives, product updates, and episodes of The Task at Hand Podcast.
- EpisodeS1 E7
Chris Griffing (GitKraken) on Where Agents Still Need Your Taste - Episode 7
Agents can write your front end, tune your SQL and build your demo app in an afternoon. In this episode, Jakob sits down with Chris Griffing, developer advocate at GitKraken and builder of Messijo, to talk about building Messijo live on Twitch, running LLMs in production, and grading models on the git commands they want to run.
- Post
Taskless v0.13.0: Smoother Setup and Shareable Reports
0.13.0 gets you from nothing to working rules with one command and one question. The setup wizard handles your agent's permissions, onboarding writes your rules, and a Taskless report shows what they find in your code today.
- Post
Deterministic Desloppification
Harry Solovay's Effect traits, turned into Taskless rules, caught all 13 planted violations. What the rules missed sorts into types, cross-file and intent, and a decision model only needs to see the intent questions once determinism cuts the list.
- Post
Taskless, ty, and the Type-Shaped Gap
Charlie Marsh's thread on ty's redundant-condition check got us asking how much a generic tool can cover. Rebuilt as Taskless rules, it reaches 13 of the thread's 15 examples, and the last two need real type inference.
- Post
Taskless v0.12.0: Revisions, Restores, and Rollbacks
0.12.0 moves the CLI onto version 2 of the rule API. check verifies every issued rule, three new commands restore, roll back and list a rule's revisions, and everything announced for 0.11.3 ships with it.
- EpisodeS1 E6
CL Kao (Recce) on What Breaks When AI Touches Your Data - Episode 6
Agents are writing the SQL that reshapes your data, and a code review only shows the change to the query. In this episode, Jakob sits down with CL Kao, founder of Recce and creator of Spacedock, to talk about data review, blast radius, and deciding when a human should look at what an agent did.
- Post
Taskless v0.11.3: Rules, Renames, and Rendered Messages
0.11.3 is a patch release with four changes you'll notice: check --rule runs a single rule, rule ids are unique across engines, Vale 3.22.0 arrives with a config schema that catches rules switching each other off, and test shows the message a rule renders.
- Post
Taskless v0.11.2: Parse Errors, Update, and Feedback
0.11.2 is a patch release with three changes you'll notice: check no longer loses every finding to one bad file, init tells you when update has work for you, and a new feedback subcommand lets you tell us whether a session went well.
- Post
Vale via npm, without an install script
Taskless ships Vale as six per-platform npm packages with no install script and nothing fetched at runtime. The pattern we copied from ast-grep, the one piece we dropped, and why npm v12 turned that accidental decision into the right one.
- Post
Your style guide was a suggestion
Taskless 0.11.x adds Vale as a second static-tier engine, so a correction about language becomes a rule that runs on every check. One team's banned-terms list was already in their AGENTS.md, and the first enforced run found 175 things their agents had read the instruction and ignored.
- Post
Taskless v0.11.x: Vale, Markdown, and More
Taskless CLI 0.11.x adds Vale as a second engine, runs all three engines in one pass, and routes a correction to the right one on its own. You describe the constraint and Taskless works out where it belongs.
- EpisodeS1 E5
Brian Douglas (Paper Compute) on Who Watches the Agents? - Episode 5
Agents are writing more of the code and leaving less of a trail behind them. In this episode, Jakob sits down with Brian Douglas, co-founder of Paper Compute, to explore why the network layer is the only place you can see what your agent actually did.
- EpisodeS1 E4
Craig Campbell (Pastmaps) on the Google Maps of History - Episode 4
What if you could slide back in time and see exactly what used to be where you're standing? That's the magic behind Pastmaps — and in this episode, Jakob Heuser sits down with Craig Campbell, the solo founder behind it.
- Post
The Security Model Behind Runtime Rules
A runtime rule is generated code that runs against your codebase. Here's how Taskless earns the trust to run it: a least-trust generation sandbox, cryptographic signing, and verification before anything runs on your machine.
- Post
Introducing Runtime Rules
Taskless 0.10.0 adds runtime rules, a new kind of rule that runs against your code to catch corrections a static pattern can't express: cross-file invariants, code checked against config, and more.
- EpisodeS1 E3
Greg Pstrucha on Practical AI - Episode 3
Agent skills are dependencies nobody locks down. Most teams install them without reading a line. In this episode, Jakob sits down with Greg Pstrucha, an engineer on Sentry's AI/ML team, to explore the attack vectors hiding in agent skills, and what it takes to treat them like real dependencies.
- Post
Taskless Now Works With the Tools You Already Have
Starting with Taskless CLI v0.9.0, Taskless skills work with your repo, find the tools you're already using, and helps you write better static analysis rules in more ways.
- EpisodeS1 E2
Jitpal Kocher on why Context Rules AI - Episode 2
Context is the missing primitive in the AI stack. Most teams are underestimating it. In this episode, Jakob sits down with JP, founder of Wire, to explore why context engineering is the foundational layer that makes agents actually work.
- EpisodeS1 E1
Vy Nguyen on Founding Taskless - Episode 1
Vy Nguyen (CEO, Taskless) sits down with her cofounder Jakob to talk about the state of AI tooling, and why we need deterministic tools more than ever.
- Post
Introducing Taskless
Tell AI Once. Stop repeating yourself to code review bots.