How Tech Signal Six works
Tech Signal Six tracks technology themes rather than individual news stories. Individual pieces of information — a repository growing quickly, a package's downloads jumping, several companies shipping products around one concept — are collected as signals, then grouped into themes. A theme stays on the platform over time while its Six Traction, momentum and maturity are updated weekly. Every weekly edition is reviewed by a human editor before publication.
Six Traction
Each theme receives a score between 0 and 100 representing how much meaningful momentum the technology currently has — not how much press coverage it gets. The score blends five components, listed here from most to least influential:
| Component | What it measures |
|---|---|
| Adoption | Repository usage, package downloads, stars, forks, contributors and enterprise implementations. |
| Momentum | How quickly adoption indicators are changing over recent weeks. |
| Source breadth | Whether the trend appears across several independent sources — developers, open source, research, cloud providers, enterprises and media. |
| Enterprise relevance | Potential relevance for organisations: productivity, infrastructure, security, cost and strategy. |
| Learning value | Whether understanding the technology is likely to remain useful beyond the immediate news cycle. |
Measured evidence of adoption and momentum carries most of the weight; editorial judgement carries the least. The weights are fixed, applied identically to every theme, and changed only through a recalibration (see below), never to move one theme up or down.
Momentum
- Accelerating
- Six Traction is increasing rapidly.
- Growing
- Six Traction is increasing steadily.
- Stable
- No significant change.
- Declining
- Interest or adoption appears to be decreasing.
Momentum is calculated from changes in underlying data, not editorial opinion.
Maturity
- Experimental
- Mostly research, prototypes or early developer experimentation.
- Emerging
- Growing ecosystem and developer interest.
- Adopt
- Increasing real-world implementation and production usage.
- Mainstream
- Widely adopted and established.
Maturity is independent of Six Traction: a theme can have a high Six Traction yet be experimental, or a moderate one yet be mainstream.
What Tech Signal Six monitors
Tech Signal Six watches 56 technologies each week: a watchlist of candidates plus the themes already featured. Only those crossing the weekly attention threshold become themes, which is why the homepage shows a handful rather than the whole list. The featured set is the output of that filter, not the extent of what is tracked.
The watchlist is not fixed. Every week the platform also looks beyond it — at fast-growing open-source projects, active research and developer discussion — and groups the strongest findings into candidate themes. Their projects are then measured week by week. A candidate becomes a theme only when it holds up: sustained over several weeks, outside every existing theme, reported by more than one independent kind of source, and backed by measurable growth. A fast-rising project inside an existing theme's area is added to that theme instead. Every change is approved by the publisher with the weekly edition, and a new theme appears only once it has been measured and its text and learning path are complete.
Evidence comes from four categories of source, collected automatically: open source (repository activity and contributors), packages (usage and reliance across the major package ecosystems), research (new papers) and community (developer discussion). Source breadth rewards themes that show up across several of these, not just one.
No single data point can move a rating on its own: one viral post, one exceptionally large package or one unusually good week counts for no more than the rules allow, and projects are matched by verified links, never by name alone.
On the radar: candidate themes
What landscape discovery found as of 2026-W43. These are not themes and carry no score; they are areas being measured to see whether they hold up.
- Fast Typed Decision Models · seen 1 week since 2026-W43 · not covered by a theme yet
Several independent projects now use small, non-autoregressive 'decision model' checkpoints that score or classify in one pass instead of generating text, for routing, trading and agent control flow.
- Browser & Computer-Use Automation Agents · seen 1 week since 2026-W43 · not covered by a theme yet
AI agents that directly operate a browser, phone or desktop are appearing across independent projects, from headless cloud browsers to phone-control bots and Android test automation.
- DeepSeek Harness Plugin Ecosystem · seen 1 week since 2026-W43 · part of AI Coding Agents
DeepSeek's new agent harness ships a plugin architecture, and within a week independent teams built marketplaces, desktop apps, routing layers and a TUI on top of it.
- Agent Skills Marketplaces · seen 1 week since 2026-W43 · part of AI Coding Agents
Coding agents such as Claude Code, Codex and Gemini CLI are converging on a shared 'skill' packaging format, and dozens of independent skill repositories appeared within a single week.
- Agent Runtime & Sandboxing Infrastructure · seen 1 week since 2026-W43 · not covered by a theme yet
Cloud vendors are shipping dedicated compute, gateways and sandboxes for running fleets of autonomous AI agents, as a compromised LLM proxy package shows the new attack surface.
Measured, derived and editorial
Not every component can be measured, so each one is labelled by how it was produced. The labels appear next to the numbers on every theme page.
- Measured
- Computed from data collected this week across open source, package ecosystems, research and developer communities.
- Derived
- Computed from measured values, or set to a neutral default when the comparison needed to measure it does not exist yet.
- Editorial
- A human judgement. Reviewed weekly, snapped to defined levels, and never presented as a measurement.
Adoption and source breadth are measured. Momentum is measured once two consecutive weekly collections exist, and neutral until then. Enterprise relevance and learning value are editorial and always labelled as such, and together they are the smaller share of every score.
How adoption is measured
Adoption combines several independent families of evidence — interest, active development, real-world usage and reliance by other software — so that closely related numbers cannot double-count. Each measure is placed on a fixed logarithmic scale, so the largest projects do not drown out everything else, and the reference points are the same for every theme. Different package ecosystems are never added together, because a download in one is not the same unit as an install in another, and a theme with no packages is not scored as though it had zero usage. Where a source cannot be measured reliably, it is left out rather than guessed.
Momentum looks at how those measures change over recent weeks, with limits that stop any single package or source from dominating the week-to-week change.
When the method or a theme's tracked sources change, the theme is recalibrated the same week: its score is computed under both the old and the new rules from the same week's data, and every earlier point moves by that difference. The weekly change then compares like with like, and each shifted point keeps its original value for audit.
Enterprise relevance and learning value snap to a small set of defined levels, because a value like 88 would imply a precision that an editorial judgement does not have.
Measured history versus editorial context
Each theme page states the date measurement began. Anything earlier is an editorial estimate of how the theme developed, kept for context in a separate, clearly labelled section — it is never drawn as one continuous line with measured points, and it never feeds a score. Weekly change is only reported as a number when both endpoints are measured; otherwise the theme reads “newly tracked” rather than showing a difference between two opinions.
The anti-hype mechanism
The ranking rewards actual adoption, developer activity, independent sources, sustained momentum and real-world implementation. It reduces the weight of repeated press coverage, company marketing, duplicate announcements and short-lived social media spikes. A technology with heavy media coverage but little measurable adoption receives a low Six Traction.
Resource freshness
Learning recommendations are at most six months old, except continuously-maintained resources (official documentation, active repositories), which are marked “maintained” and re-verified during weekly review.
The weekly cycle
- 1.Collect new data from open source, package ecosystems, research, vendors and foundations.
- 2.Detect technologies with unusual increases or decreases in activity.
- 3.Cluster related signals into technology themes.
- 4.Calculate Six Traction, momentum and maturity.
- 5.Identify new, rising, falling and newsworthy themes.
- 6.Generate draft summaries and learning recommendations.
- 7.Human editorial review of themes, descriptions, scores and resources.
- 8.Publish the new weekly edition.