How Tech Signal works
Tech Signal 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 traction, momentum and maturity are updated weekly. Every weekly edition is reviewed by a human editor before publication.
The Traction Score
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 is a weighted blend of five components:
| Component | Weight | What it measures |
|---|---|---|
| Adoption | 30% | Repository usage, package downloads, stars, forks, contributors and enterprise implementations. |
| Momentum | 25% | How quickly adoption indicators are changing (4- and 8-week growth). |
| Source breadth | 20% | Whether the trend appears across several independent sources — developers, open source, research, cloud providers, enterprises and media. |
| Enterprise relevance | 15% | Potential relevance for organisations: productivity, infrastructure, security, cost and strategy. |
| Learning value | 10% | Whether understanding the technology is likely to remain useful beyond the immediate news cycle. |
Momentum
- Accelerating
- Traction is increasing rapidly.
- Growing
- 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 traction: a theme can be high-traction yet experimental, or moderate-traction yet mainstream.
What Tech Signal monitors
Tech Signal watches 55 technologies each week — 47 candidates on a watchlist 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.
Four categories of source are collected automatically: open source (GitHub stars, forks, contributors, commit velocity, releases), packages (npm and PyPI weekly downloads), research (arXiv submissions) and community (Hacker News discussion). Source breadth scores how many of those four report on a theme: 25 points plus 16 for each category.
Hacker News points are never summed, and the highest-scoring story is never an input to any score — a single viral post counts exactly as much as any other story that clears the threshold. This is what stops one popular link from moving a technology's rating.
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 — GitHub repositories, package registries, arXiv submissions and Hacker News discussion.
- 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 — together they are a quarter of every score.
How adoption is anchored
Adoption maps raw counts onto a 0–100 scale using fixed reference points on a logarithmic scale. Each ecosystem is scored separately and the ecosystems that reported are averaged — npm and PyPI totals are not added together, because npm figures are inflated by CI mirrors, and a theme with no packages is not scored as though it had zero downloads.
| GitHub stars | Score | Weekly downloads | Score |
|---|---|---|---|
| 1,000 | 20 | 100,000 | 30 |
| 10,000 | 40 | 1,000,000 | 50 |
| 100,000 | 60 | 10,000,000 | 70 |
| 1,000,000 | 80 | 100,000,000 | 90 |
Enterprise relevance and learning value snap to five levels — 30 marginal, 45 situational, 60 relevant, 75 important, 90 critical — because a value like 88 would imply a precision that an editorial judgement does not have.
The 2026-08-09 recalibration
Tech Signal launched with scores set editorially, before the collection pipeline ran. Those values sat above anything the published formulas could produce, so the first real collection would have dropped every theme for reasons unrelated to the technology. On 2026-08-09 the formulas above were run against measured data and the results replaced the seeded values.
The illustrative history was shifted by a single offset per theme, so no point moved relative to its neighbours: the shape of each trajectory is unchanged and only its level moved, set by formula rather than judgement. Every snapshot keeps the score it held before, and no theme was adjusted to protect its ranking.
| Theme | Before | After | Change |
|---|---|---|---|
| AI Coding Agents | 101 | 82 | -19 |
| Agent Interoperability Protocols | 88 | 75 | -13 |
| Small Language Models & On-Device AI | 82 | 68 | -14 |
| AI Inference Infrastructure | 70 | 66 | -4 |
| Vector Search & Agent Memory | 71 | 66 | -5 |
| Memory-Safe Systems Programming | 64 | 63 | -1 |
| WebAssembly Beyond the Browser | 67 | 62 | -5 |
| Post-Quantum Cryptography | 79 | 58 | -21 |
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 Traction Score.
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 configured sources (GitHub, package registries, research, vendors, foundations).
- 2.Detect technologies with unusual increases or decreases in activity.
- 3.Cluster related signals into technology themes.
- 4.Calculate Traction Score, 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.