Resource Evaluation: AI's Impact on Software Engineering in 2026
URL: https://eventuallymaking.io/p/ai-s-impact-on-the-state-of-the-art-in-software-engineering-in-2026
Author: Hugo (Software Engineer, 20+ years, Founder Malt/Writizzy)
Published: February 6, 2026
Type: Opinion article based on 7 French company interviews
Evaluated: February 6, 2026
Evaluator: Claude Code Guide Team
Summary
Opinion piece on AI's impact on software engineering practices in 2026, based on interviews with 7 French tech companies (Doctolib, Malt, Alan, Google Cloud, Brevo, ManoMano, Ilek, Clever Cloud).
Key arguments:
- Context Engineering (Thoughtworks framework) — shift toward complete specifications with constraints before coding
- Spec/Plan/Act workflow standardization — industry consensus on 3-phase approach
- Corporate AI governance — organizational marketplaces to pool AI skills, agents, rules
- QA via CI/CD — traditional practices (linting, testing, review) essential for AI-generated code validation
- HR disruption — junior training, recruitment, career trajectories require restructuring
Stats cited:
- Monthly costs: ~$20/dev (adoption), ~$200/dev (strong adoption), $200-1000+/dev (advanced multi-agent)
- 90%+ of engineers at Alan use AI-powered coding assistants daily
- Interviews: 7 companies (no detailed verbatims provided)
Evaluation Scores
| Criterion |
Score (1-5) |
Notes |
| Relevance |
2 |
Marginal — concepts largely covered in guide, one legitimate gap (Context Engineering) |
| Accuracy |
3 |
Moderate — terminology error ("Context Driven" vs "Context Engineering"), stats lack methodology |
| Actionability |
1 |
Low — no templates, code, or concrete workflows |
| Novelty |
2 |
Marginal — Spec/Plan/Act and QA/CI/CD already in guide, Context Engineering framework new |
| Production-Ready |
1 |
Low — opinion piece, no implementation details |
Overall Score: 2/5 (Marginal - Info secondaire)
Gap Analysis
What's NEW (not in guide)
| Aspect |
Hugo's Resource |
Our Guide |
Gap? |
| Context Engineering (Thoughtworks) |
✓ Mentioned (but miscited as "Context Driven") |
✗ Absent |
✅ Legitimate gap |
| Corporate AI marketplaces |
✓ Concept described |
✗ Not covered |
⚠️ Minor gap (RH focus, not technical) |
What's ALREADY COVERED
| Aspect |
Hugo's Resource |
Our Guide |
| Spec/Plan/Act workflow |
✓ Described |
✅ guide/workflows/spec-first.md, /plan mode |
| QA via CI/CD |
✓ Mentioned |
✅ guide/production-safety.md, hooks |
| HR/Junior disruption |
✓ Opinion |
✅ guide/learning-with-ai.md (comprehensive) |
| Cost estimates |
✓ Ranges ($20-1000) |
✅ guide/ai-ecosystem.md (precise: $20-50) |
Fact-Check Results
| Claim |
Verified |
Source |
Correction |
| "Context Driven Engineering" |
⚠️ Terminology error |
Perplexity search |
✅ Correct term: "Context Engineering" (Thoughtworks Tech Radar Vol 33, Nov 2025) |
| "90%+ engineers at Alan" |
✅ Yes |
Emma Goldblum quote (article) |
✅ Verbatim exact |
| "$20-200-1000/dev costs" |
✅ Table present |
Article |
⚠️ No methodology, 50x spread too large |
| "Hugo 20+ years XP" |
✅ Yes |
Schema markup |
✅ Malt CTO 2012-2024, Writizzy founder 2025 |
| "Published Feb 6, 2026" |
✅ Yes |
Metadata |
✅ Correct |
| "Interviews 7 companies" |
✅ List present |
Article |
⚠️ No verbatims, no raw data |
Critical error detected: Hugo miscites Thoughtworks framework as "Context Driven Engineering" when the actual term is "Context Engineering" (verified via Perplexity and Thoughtworks Technology Radar Vol 33).
Technical-Writer Challenge
Agent ID: ae2f481 (technical-writer subagent)
Challenge summary:
- Initial score 4/5 reduced to 2/5 after critical analysis
- Overestimated novelty — Spec/Plan/Act already in
spec-first.md, QA/CI/CD in production-safety.md
- Underestimated marketing angle — no peer review, stats lack methodology, Writizzy link in footer
- Compared unfavorably to validated score-4 resources (Pat Cullen: 3 templates, Paddo: 10 actionable tips)
Legitimate points:
- "Context Engineering" (Thoughtworks) is a real gap in the guide
- Corporate governance angle minimally covered
- Stats too vague for practical use ($20-1000 spread, no methodology)
Recommendation upheld: Minimal integration (footnotes only), not full section.
Integration Decision
Action taken: Minimal integration (2 footnotes)
1. Context Engineering (Thoughtworks) — Priority HIGH
File: guide/methodologies.md (after line 66, "Foundational Discipline" section)
Added:
> **Context Engineering**: Thoughtworks designates this broader approach "Context Engineering"
> in their Technology Radar (Nov 2025) — the systematic design of information provided to LLMs
> during inference. Three core techniques: context setup, context management for long-horizon
> tasks, and dynamic information retrieval. Related patterns in Claude Code: AGENTS.md,
> MCP Context7, Plan Mode.
Rationale: Legitimate framework gap, verified via Perplexity and Thoughtworks documentation.
2. Corporate AI Marketplaces — Priority LOW
File: guide/adoption-approaches.md (after line 277, "Larger Team" section)
Added:
> **Emerging approach**: Some organizations explore "corporate AI marketplaces" to pool AI
> skills, agents, and rules at the organizational level rather than individual teams
> (Hugo/Writizzy 2026). Few documented production implementations yet, but the concept
> addresses governance at scale.
Rationale: Interesting RH concept, minimal technical implementation details available.
Why NOT More Integration?
Rejected: Full "Team Governance" section
Reason: Redundant with existing content:
guide/adoption-approaches.md lines 236-278 already cover team coordinationguide/production-safety.md covers hooks and permission rulesguide/security-hardening.md covers team conventions
Rejected: Stats integration
Reason: Unusable methodology:
- "$20-1000/dev" range is 50x spread
- No methodology documentation
- Our guide has more precise estimates (
ai-ecosystem.md: $20-50 Claude Code typical)
Rejected: Citing "Context Driven Engineering"
Reason: Term doesn't exist — Thoughtworks framework is "Context Engineering"
Comparison to Other Evaluations
| Resource |
Score |
Templates/Code |
Stats Quality |
Integration |
| Pat Cullen (review-pr) |
4/5 |
3 templates |
N/A |
Full guide section |
| Paddo Team Tips |
4/5 |
0 (10 actionable tips) |
N/A |
Integrated throughout |
| RTK |
4/5 |
1 tool + examples |
Measured 72.6% reduction |
Full guide section |
| Hugo AI Impact |
2/5 |
0 |
Vague ($20-1000 spread) |
2 footnotes only |
Lessons Learned
Evaluation Process Improvements
- Terminology verification: Always cross-check framework names with authoritative sources (Perplexity, official docs)
- Gap analysis rigor: Grep existing guide before claiming "missing content"
- Stats scrutiny: Require methodology documentation, not just numbers
- Technical-writer challenge: Proved valuable — caught overestimation of novelty
What Worked
- Fact-check protocol: Caught terminology error early
- Agent challenge: technical-writer agent provided brutal but accurate reality check
- Minimal integration: 10-minute footnotes vs 2-hour full section = better ROI
Sources
Metadata
- Evaluation date: 2026-02-06
- Time spent: ~45 minutes (research, fact-check, agent challenge, integration)
- Agent tools used: WebFetch, Grep, Read, Perplexity, Task (technical-writer)
- Integration time: 10 minutes (2 footnotes)
- Files modified: 2 (
guide/methodologies.md, guide/adoption-approaches.md)
1---2name: resource-evaluation-ai-s-impact-on-software-engineering-in3description: Opinion piece on AI's impact on software engineering practices in 2026, based on interviews with 7 French tech companies (Doctolib, Malt, Alan, Google Cloud, Brevo, ManoMano, Ilek, Clever Cloud).4---5# Resource Evaluation: AI's Impact on Software Engineering in 202667**URL**: https://eventuallymaking.io/p/ai-s-impact-on-the-state-of-the-art-in-software-engineering-in-20268**Author**: Hugo (Software Engineer, 20+ years, Founder Malt/Writizzy)9**Published**: February 6, 202610**Type**: Opinion article based on 7 French company interviews11**Evaluated**: February 6, 202612**Evaluator**: Claude Code Guide Team1314---1516## Summary1718Opinion piece on AI's impact on software engineering practices in 2026, based on interviews with 7 French tech companies (Doctolib, Malt, Alan, Google Cloud, Brevo, ManoMano, Ilek, Clever Cloud).1920**Key arguments**:211. **Context Engineering** (Thoughtworks framework) — shift toward complete specifications with constraints before coding222. **Spec/Plan/Act workflow standardization** — industry consensus on 3-phase approach233. **Corporate AI governance** — organizational marketplaces to pool AI skills, agents, rules244. **QA via CI/CD** — traditional practices (linting, testing, review) essential for AI-generated code validation255. **HR disruption** — junior training, recruitment, career trajectories require restructuring2627**Stats cited**:28- Monthly costs: ~$20/dev (adoption), ~$200/dev (strong adoption), $200-1000+/dev (advanced multi-agent)29- 90%+ of engineers at Alan use AI-powered coding assistants daily30- Interviews: 7 companies (no detailed verbatims provided)3132---3334## Evaluation Scores3536| Criterion | Score (1-5) | Notes |37|-----------|-------------|-------|38| **Relevance** | 2 | Marginal — concepts largely covered in guide, one legitimate gap (Context Engineering) |39| **Accuracy** | 3 | Moderate — terminology error ("Context Driven" vs "Context Engineering"), stats lack methodology |40| **Actionability** | 1 | Low — no templates, code, or concrete workflows |41| **Novelty** | 2 | Marginal — Spec/Plan/Act and QA/CI/CD already in guide, Context Engineering framework new |42| **Production-Ready** | 1 | Low — opinion piece, no implementation details |4344**Overall Score**: **2/5** (Marginal - Info secondaire)4546---4748## Gap Analysis4950### What's NEW (not in guide)5152| Aspect | Hugo's Resource | Our Guide | Gap? |53|--------|----------------|-----------|------|54| **Context Engineering** (Thoughtworks) | ✓ Mentioned (but miscited as "Context Driven") | ✗ Absent | ✅ **Legitimate gap** |55| **Corporate AI marketplaces** | ✓ Concept described | ✗ Not covered | ⚠️ **Minor gap** (RH focus, not technical) |5657### What's ALREADY COVERED5859| Aspect | Hugo's Resource | Our Guide |60|--------|----------------|-----------|61| Spec/Plan/Act workflow | ✓ Described | ✅ `guide/workflows/spec-first.md`, `/plan` mode |62| QA via CI/CD | ✓ Mentioned | ✅ `guide/production-safety.md`, hooks |63| HR/Junior disruption | ✓ Opinion | ✅ `guide/learning-with-ai.md` (comprehensive) |64| Cost estimates | ✓ Ranges ($20-1000) | ✅ `guide/ai-ecosystem.md` (precise: $20-50) |6566---6768## Fact-Check Results6970| Claim | Verified | Source | Correction |71|-------|----------|--------|------------|72| **"Context Driven Engineering"** | ⚠️ **Terminology error** | Perplexity search | ✅ Correct term: "Context Engineering" (Thoughtworks Tech Radar Vol 33, Nov 2025) |73| **"90%+ engineers at Alan"** | ✅ Yes | Emma Goldblum quote (article) | ✅ Verbatim exact |74| **"$20-200-1000/dev costs"** | ✅ Table present | Article | ⚠️ No methodology, 50x spread too large |75| **"Hugo 20+ years XP"** | ✅ Yes | Schema markup | ✅ Malt CTO 2012-2024, Writizzy founder 2025 |76| **"Published Feb 6, 2026"** | ✅ Yes | Metadata | ✅ Correct |77| **"Interviews 7 companies"** | ✅ List present | Article | ⚠️ No verbatims, no raw data |7879**Critical error detected**: Hugo miscites Thoughtworks framework as "Context Driven Engineering" when the actual term is "Context Engineering" (verified via Perplexity and Thoughtworks Technology Radar Vol 33).8081---8283## Technical-Writer Challenge8485**Agent ID**: `ae2f481` (technical-writer subagent)8687**Challenge summary**:88- Initial score 4/5 **reduced to 2/5** after critical analysis89- Overestimated novelty — Spec/Plan/Act already in `spec-first.md`, QA/CI/CD in `production-safety.md`90- Underestimated marketing angle — no peer review, stats lack methodology, Writizzy link in footer91- Compared unfavorably to validated score-4 resources (Pat Cullen: 3 templates, Paddo: 10 actionable tips)9293**Legitimate points**:94- "Context Engineering" (Thoughtworks) is a real gap in the guide95- Corporate governance angle minimally covered96- Stats too vague for practical use ($20-1000 spread, no methodology)9798**Recommendation upheld**: Minimal integration (footnotes only), not full section.99100---101102## Integration Decision103104**Action taken**: **Minimal integration** (2 footnotes)105106### 1. Context Engineering (Thoughtworks) — Priority HIGH107108**File**: `guide/methodologies.md` (after line 66, "Foundational Discipline" section)109110**Added**:111```markdown112> **Context Engineering**: Thoughtworks designates this broader approach "Context Engineering"113> in their Technology Radar (Nov 2025) — the systematic design of information provided to LLMs114> during inference. Three core techniques: context setup, context management for long-horizon115> tasks, and dynamic information retrieval. Related patterns in Claude Code: AGENTS.md,116> MCP Context7, Plan Mode.117```118119**Rationale**: Legitimate framework gap, verified via Perplexity and Thoughtworks documentation.120121### 2. Corporate AI Marketplaces — Priority LOW122123**File**: `guide/adoption-approaches.md` (after line 277, "Larger Team" section)124125**Added**:126```markdown127> **Emerging approach**: Some organizations explore "corporate AI marketplaces" to pool AI128> skills, agents, and rules at the organizational level rather than individual teams129> (Hugo/Writizzy 2026). Few documented production implementations yet, but the concept130> addresses governance at scale.131```132133**Rationale**: Interesting RH concept, minimal technical implementation details available.134135---136137## Why NOT More Integration?138139### Rejected: Full "Team Governance" section140141**Reason**: Redundant with existing content:142- `guide/adoption-approaches.md` lines 236-278 already cover team coordination143- `guide/production-safety.md` covers hooks and permission rules144- `guide/security-hardening.md` covers team conventions145146### Rejected: Stats integration147148**Reason**: Unusable methodology:149- "$20-1000/dev" range is 50x spread150- No methodology documentation151- Our guide has more precise estimates (`ai-ecosystem.md`: $20-50 Claude Code typical)152153### Rejected: Citing "Context Driven Engineering"154155**Reason**: Term doesn't exist — Thoughtworks framework is "Context Engineering"156157---158159## Comparison to Other Evaluations160161| Resource | Score | Templates/Code | Stats Quality | Integration |162|----------|-------|----------------|---------------|-------------|163| **Pat Cullen** (review-pr) | 4/5 | 3 templates | N/A | Full guide section |164| **Paddo Team Tips** | 4/5 | 0 (10 actionable tips) | N/A | Integrated throughout |165| **RTK** | 4/5 | 1 tool + examples | Measured 72.6% reduction | Full guide section |166| **Hugo AI Impact** | **2/5** | 0 | Vague ($20-1000 spread) | **2 footnotes only** |167168---169170## Lessons Learned171172### Evaluation Process Improvements1731741. **Terminology verification**: Always cross-check framework names with authoritative sources (Perplexity, official docs)1752. **Gap analysis rigor**: Grep existing guide before claiming "missing content"1763. **Stats scrutiny**: Require methodology documentation, not just numbers1774. **Technical-writer challenge**: Proved valuable — caught overestimation of novelty178179### What Worked1801811. **Fact-check protocol**: Caught terminology error early1822. **Agent challenge**: technical-writer agent provided brutal but accurate reality check1833. **Minimal integration**: 10-minute footnotes vs 2-hour full section = better ROI184185---186187## Sources188189- **Primary**: Hugo, ["AI's Impact on State of the Art in Software Engineering in 2026"](https://eventuallymaking.io/p/ai-s-impact-on-the-state-of-the-art-in-software-engineering-in-2026), Feb 6, 2026190- **Verification**: Perplexity search for "Context Engineering Thoughtworks 2024 2025"191- **Authoritative**: [Thoughtworks Technology Radar Vol 33](https://www.thoughtworks.com/content/dam/thoughtworks/documents/radar/2025/11/tr_technology_radar_vol_33_en.pdf), Nov 2025192- **Supporting**: [Thoughtworks macro trends blog](https://www.thoughtworks.com/insights/blog/technology-strategy/macro-trends-tech-industry-november-2025)193194---195196## Metadata197198- **Evaluation date**: 2026-02-06199- **Time spent**: ~45 minutes (research, fact-check, agent challenge, integration)200- **Agent tools used**: WebFetch, Grep, Read, Perplexity, Task (technical-writer)201- **Integration time**: 10 minutes (2 footnotes)202- **Files modified**: 2 (`guide/methodologies.md`, `guide/adoption-approaches.md`)
tools-only